[llvm] [True16] Make loads into s16 legal in GlobalISel (PR #176963)
Domenic Nutile via llvm-commits
llvm-commits at lists.llvm.org
Mon Feb 2 07:56:10 PST 2026
https://github.com/saxlungs updated https://github.com/llvm/llvm-project/pull/176963
>From 546f77ee138a7d41ad31dd9718e8e0d75bb2266e Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Tue, 20 Jan 2026 11:34:33 -0500
Subject: [PATCH 1/2] [True16] Make loads into s16 legal in GlobalISel
Enable selection of the FLAT, GLOBAL, and SCRATCH True16 D16 load stores
---
.../AMDGPU/AMDGPUInstructionSelector.cpp | 2 +-
.../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 27 +-
.../CodeGen/AMDGPU/GlobalISel/flat-scratch.ll | 81 +-
llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll | 52 +-
llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll | 65 +-
.../AMDGPU/GlobalISel/insertelement.i8.ll | 499 +-
.../AMDGPU/GlobalISel/insertelement.ll | 2 +-
.../AMDGPU/GlobalISel/legalize-fdiv.mir | 84 +-
.../GlobalISel/legalize-load-constant.mir | 4631 ++++-
.../AMDGPU/GlobalISel/legalize-load-flat.mir | 3368 +++-
.../GlobalISel/legalize-load-global.mir | 14378 +++++++++++++++-
.../AMDGPU/GlobalISel/legalize-load-local.mir | 3446 +++-
.../GlobalISel/legalize-load-private.mir | 1106 +-
.../GlobalISel/legalize-store-global.mir | 3568 +++-
.../AMDGPU/GlobalISel/store-local.128.ll | 64 +-
.../AMDGPU/GlobalISel/store-local.96.ll | 48 +-
.../test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll | 284 +-
.../test/CodeGen/AMDGPU/GlobalISel/usubsat.ll | 284 +-
.../build-vector-packed-partial-undef.ll | 54 +-
.../CodeGen/AMDGPU/dagcombine-fmul-sel.ll | 10 +-
llvm/test/CodeGen/AMDGPU/fma.f16.ll | 20 +-
llvm/test/CodeGen/AMDGPU/fmed3.ll | 23 +-
llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll | 45 +-
llvm/test/CodeGen/AMDGPU/global-saddr-load.ll | 114 +-
.../AMDGPU/llvm.amdgcn.intersect_ray.ll | 121 +-
llvm/test/CodeGen/AMDGPU/llvm.frexp.ll | 44 +-
.../CodeGen/AMDGPU/llvm.is.fpclass.f16.ll | 37 +-
llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll | 37 +-
llvm/test/CodeGen/AMDGPU/llvm.log.ll | 67 +-
llvm/test/CodeGen/AMDGPU/llvm.log10.ll | 67 +-
llvm/test/CodeGen/AMDGPU/llvm.log2.ll | 45 +-
llvm/test/CodeGen/AMDGPU/roundeven.ll | 12 +-
llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll | 12 +-
llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll | 46 +-
llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll | 26 +-
.../test/CodeGen/AMDGPU/vector-reduce-fadd.ll | 6 +-
.../test/CodeGen/AMDGPU/vector-reduce-fmax.ll | 142 +-
.../test/CodeGen/AMDGPU/vector-reduce-fmin.ll | 142 +-
.../CodeGen/AMDGPU/vector-reduce-fminimum.ll | 54 +-
.../test/CodeGen/AMDGPU/vector-reduce-fmul.ll | 6 +-
40 files changed, 31153 insertions(+), 1966 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
index e239e6f56cb44..41f8773ff0e7e 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -665,7 +665,7 @@ bool AMDGPUInstructionSelector::selectG_MERGE_VALUES(MachineInstr &MI) const {
LLT SrcTy = MRI->getType(MI.getOperand(1).getReg());
const unsigned SrcSize = SrcTy.getSizeInBits();
- if (SrcSize < 32)
+ if (SrcSize < 16)
return selectImpl(MI, *CoverageInfo);
const DebugLoc &DL = MI.getDebugLoc();
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 2b2046fb9f65b..b30c676622768 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -462,8 +462,14 @@ static bool isLoadStoreSizeLegal(const GCNSubtarget &ST,
MemSize = std::max(MemSize, Align);
#endif
- // Only 1-byte and 2-byte to 32-bit extloads are valid.
- if (MemSize != RegSize && RegSize != 32)
+ // Only 8-bit and 16-bit to 32-bit extloads are valid.
+ if (!ST.useRealTrue16Insts() && MemSize != RegSize && RegSize != 32)
+ return false;
+
+ // 8-bit to 16-bit extloads are valid using True16 instructions.
+ // 8-bit and 16-bit to 32-bit extloads are valid for non-True16 D16 memory
+ // ops, which could be re-implemented as True16
+ if (ST.useRealTrue16Insts() && MemSize != RegSize && RegSize > 32)
return false;
if (MemSize > maxSizeForAddrSpace(ST, AS, IsLoad,
@@ -542,7 +548,7 @@ static bool loadStoreBitcastWorkaround(const LLT Ty) {
static bool isLoadStoreLegal(const GCNSubtarget &ST, const LegalityQuery &Query) {
const LLT Ty = Query.Types[0];
- return isRegisterType(ST, Ty) && isLoadStoreSizeLegal(ST, Query) &&
+ return isRegisterClassType(ST, Ty) && isLoadStoreSizeLegal(ST, Query) &&
!hasBufferRsrcWorkaround(Ty) && !loadStoreBitcastWorkaround(Ty);
}
@@ -1549,6 +1555,15 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
{V4S32, ConstantPtr, V4S32, GlobalAlign32},
{S64, ConstantPtr, S64, GlobalAlign32},
{V2S32, ConstantPtr, V2S32, GlobalAlign32}});
+
+ if (ST.useRealTrue16Insts())
+ Actions.legalForTypesWithMemDesc({{S16, GlobalPtr, S8, GlobalAlign8},
+ {S16, GlobalPtr, S16, GlobalAlign16},
+ {S16, LocalPtr, S8, 8},
+ {S16, LocalPtr, S16, 16},
+ {S16, PrivatePtr, S8, 8},
+ {S16, PrivatePtr, S16, 16}});
+
Actions.legalIf(
[=](const LegalityQuery &Query) -> bool {
return isLoadStoreLegal(ST, Query);
@@ -2023,7 +2038,6 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
auto &Builder =
getActionDefinitionsBuilder(Op)
- .legalIf(all(isRegisterType(ST, 0), isRegisterType(ST, 1)))
.lowerFor({{S16, V2S16}})
.lowerIf([=](const LegalityQuery &Query) {
const LLT BigTy = Query.Types[BigTyIdx];
@@ -2054,7 +2068,10 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
return notValidElt(Query, BigTyIdx);
},
scalarize(1))
- .clampScalar(BigTyIdx, S32, MaxScalar);
+ .clampScalar(BigTyIdx, S32, MaxScalar)
+ // legalIf must come after other rules because we don't have 16-bit SGPRs.
+ // We expect 16-bit values to already be lowered or widened by this point.
+ .legalIf(all(isRegisterType(ST, 0), isRegisterType(ST, 1)));
if (Op == G_MERGE_VALUES) {
Builder.widenScalarIf(
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
index 13d6597bd78f1..d83073fa1f256 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
@@ -2695,22 +2695,23 @@ define void @store_load_i64_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX11-LABEL: store_load_i64_unaligned:
; UNALIGNED_GFX11: ; %bb.0: ; %bb
; UNALIGNED_GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; UNALIGNED_GFX11-NEXT: v_dual_mov_b32 v1, 15 :: v_dual_mov_b32 v2, 0
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.l, 15
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.h, 0
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:1 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:2 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:3 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:4 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:4 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:5 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:6 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:7 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; UNALIGNED_GFX11-NEXT: scratch_load_u8 v1, v0, off glc dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt vmcnt(0)
@@ -3034,32 +3035,33 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX11-LABEL: store_load_v3i32_unaligned:
; UNALIGNED_GFX11: ; %bb.0: ; %bb
; UNALIGNED_GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; UNALIGNED_GFX11-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
-; UNALIGNED_GFX11-NEXT: v_mov_b32_e32 v3, 2
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.l, 1
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v2.l, 2
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:1 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:2 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:3 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: v_mov_b32_e32 v1, 3
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v3, off offset:4 dlc
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.l, 3
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:4 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:5 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:6 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:7 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off offset:8 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:9 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:10 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:11 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; UNALIGNED_GFX11-NEXT: scratch_load_u8 v1, v0, off glc dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt vmcnt(0)
@@ -3464,41 +3466,42 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX11-LABEL: store_load_v4i32_unaligned:
; UNALIGNED_GFX11: ; %bb.0: ; %bb
; UNALIGNED_GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; UNALIGNED_GFX11-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
-; UNALIGNED_GFX11-NEXT: v_mov_b32_e32 v3, 2
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.l, 1
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v2.l, 2
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:1 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:2 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:3 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v3, off offset:4 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:4 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: v_mov_b32_e32 v1, 3
-; UNALIGNED_GFX11-NEXT: v_mov_b32_e32 v3, 4
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:5 dlc
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.l, 3
+; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v2.l, 4
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:6 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:7 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off offset:8 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:9 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:10 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:11 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v3, off offset:12 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:12 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:13 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:13 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:14 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:14 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:15 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:15 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; UNALIGNED_GFX11-NEXT: scratch_load_u8 v1, v0, off glc dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll
index 990153ebce243..c6096e5249364 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll
@@ -137,34 +137,20 @@ define amdgpu_ps void @fpext_f16_to_f64_div(half %a, ptr addrspace(1) %ptr) {
}
define amdgpu_ps <2 x float> @fpext_v2f16_to_v2f32_uniform(<2 x half> inreg %a) {
-; GFX11-FAKE16-LABEL: fpext_v2f16_to_v2f32_uniform:
-; GFX11-FAKE16: ; %bb.0:
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 16
-; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v0, s0
-; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v1, s1
-; GFX11-FAKE16-NEXT: ; return to shader part epilog
-;
-; GFX11-TRUE16-LABEL: fpext_v2f16_to_v2f32_uniform:
-; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, s0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v0
-; GFX11-TRUE16-NEXT: ; return to shader part epilog
-;
-; GFX12-FAKE16-LABEL: fpext_v2f16_to_v2f32_uniform:
-; GFX12-FAKE16: ; %bb.0:
-; GFX12-FAKE16-NEXT: s_cvt_f32_f16 s1, s0
-; GFX12-FAKE16-NEXT: s_cvt_hi_f32_f16 s0, s0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v0, s1 :: v_dual_mov_b32 v1, s0
-; GFX12-FAKE16-NEXT: ; return to shader part epilog
+; GFX11-LABEL: fpext_v2f16_to_v2f32_uniform:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_lshr_b32 s1, s0, 16
+; GFX11-NEXT: v_cvt_f32_f16_e32 v0, s0
+; GFX11-NEXT: v_cvt_f32_f16_e32 v1, s1
+; GFX11-NEXT: ; return to shader part epilog
;
-; GFX12-TRUE16-LABEL: fpext_v2f16_to_v2f32_uniform:
-; GFX12-TRUE16: ; %bb.0:
-; GFX12-TRUE16-NEXT: s_cvt_f32_f16 s0, s0
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s0
-; GFX12-TRUE16-NEXT: ; return to shader part epilog
+; GFX12-LABEL: fpext_v2f16_to_v2f32_uniform:
+; GFX12: ; %bb.0:
+; GFX12-NEXT: s_cvt_f32_f16 s1, s0
+; GFX12-NEXT: s_cvt_hi_f32_f16 s0, s0
+; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
+; GFX12-NEXT: v_dual_mov_b32 v0, s1 :: v_dual_mov_b32 v1, s0
+; GFX12-NEXT: ; return to shader part epilog
%result = fpext <2 x half> %a to <2 x float>
ret <2 x float> %result
}
@@ -180,10 +166,10 @@ define amdgpu_ps <2 x float> @fpext_v2f16_to_v2f32_div(<2 x half> %a) {
;
; GFX11-TRUE16-LABEL: fpext_v2f16_to_v2f32_div:
; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v0.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v0.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
; GFX11-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-FAKE16-LABEL: fpext_v2f16_to_v2f32_div:
@@ -196,10 +182,10 @@ define amdgpu_ps <2 x float> @fpext_v2f16_to_v2f32_div(<2 x half> %a) {
;
; GFX12-TRUE16-LABEL: fpext_v2f16_to_v2f32_div:
; GFX12-TRUE16: ; %bb.0:
-; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v0.l
-; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v0.h
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.l
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v0, v2
+; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
; GFX12-TRUE16-NEXT: ; return to shader part epilog
%result = fpext <2 x half> %a to <2 x float>
ret <2 x float> %result
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
index b19216ae7bbc5..c3c0ebaaf69d4 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
@@ -496,18 +496,21 @@ define <2 x half> @v_pow_v2f16(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16-LABEL: v_pow_v2f16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
@@ -640,19 +643,21 @@ define <2 x half> @v_pow_v2f16_fneg_lhs(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
@@ -787,20 +792,22 @@ define <2 x half> @v_pow_v2f16_fneg_rhs(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16-LABEL: v_pow_v2f16_fneg_rhs:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
@@ -940,20 +947,22 @@ define <2 x half> @v_pow_v2f16_fneg_lhs_rhs(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll
index 7c212f1e110d1..b8d6cf57aed34 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll
@@ -3,7 +3,8 @@
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji < %s | FileCheck --check-prefix=GFX8 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=hawaii < %s | FileCheck --check-prefix=GFX7 %s
; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck --check-prefix=GFX10 %s
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX11 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck --check-prefixes=GFX11,GFX11-TRUE16 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX11,GFX11-FAKE16 %s
define amdgpu_ps void @insertelement_s_v2i8_s_s(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 inreg %idx) {
; GFX9-LABEL: insertelement_s_v2i8_s_s:
@@ -85,27 +86,49 @@ define amdgpu_ps void @insertelement_s_v2i8_s_s(ptr addrspace(4) inreg %ptr, i8
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_s_v2i8_s_s:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_cmp_eq_u32_e64 s0, s5, 1
-; GFX11-NEXT: global_load_u16 v0, v0, s[2:3]
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s4, s0
-; GFX11-NEXT: v_cmp_eq_u32_e64 s0, s5, 0
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, s4, s0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_s_v2i8_s_s:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 s0, s5, 1
+; GFX11-TRUE16-NEXT: global_load_u16 v0, v0, s[2:3]
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v1, v1, s4, s0
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 s0, s5, 0
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v2, v0, s4, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_s_v2i8_s_s:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 s0, s5, 1
+; GFX11-FAKE16-NEXT: global_load_u16 v0, v0, s[2:3]
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v1, s4, s0
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 s0, s5, 0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v0, s4, s0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(4) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
@@ -188,26 +211,47 @@ define amdgpu_ps void @insertelement_v_v2i8_s_s(ptr addrspace(1) %ptr, i8 inreg
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_v_v2i8_s_s:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: global_load_u16 v0, v[0:1], off
-; GFX11-NEXT: v_cmp_eq_u32_e64 s0, s3, 1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s2, s0
-; GFX11-NEXT: v_cmp_eq_u32_e64 s0, s3, 0
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, s2, s0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_v_v2i8_s_s:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 s0, s3, 1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v1, v1, s2, s0
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 s0, s3, 0
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v2, v0, s2, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_v_v2i8_s_s:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 s0, s3, 1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v1, s2, s0
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 s0, s3, 0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v0, s2, s0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(1 ) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
@@ -291,26 +335,48 @@ define amdgpu_ps void @insertelement_s_v2i8_v_s(ptr addrspace(4) inreg %ptr, i8
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_s_v2i8_v_s:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: global_load_u16 v1, v1, s[2:3]
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v2, 8, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 0
-; GFX11-NEXT: v_dual_cndmask_b32 v0, v1, v0 :: v_dual_and_b32 v1, 0xff, v2
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_s_v2i8_v_s:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: global_load_u16 v1, v1, s[2:3]
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v2.l, 8, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 0
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v1, v0, vcc_lo
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_s_v2i8_v_s:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: global_load_u16 v1, v1, s[2:3]
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 8, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s4, 0
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v0, v1, v0 :: v_dual_and_b32 v1, 0xff, v2
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(4) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
@@ -397,27 +463,49 @@ define amdgpu_ps void @insertelement_s_v2i8_s_v(ptr addrspace(4) inreg %ptr, i8
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_s_v2i8_s_v:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
-; GFX11-NEXT: global_load_u16 v1, v1, s[2:3]
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v2, 8, v1
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, s4, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-NEXT: v_cndmask_b32_e64 v0, v1, s4, vcc_lo
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v2
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_s_v2i8_s_v:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
+; GFX11-TRUE16-NEXT: global_load_u16 v1, v1, s[2:3]
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v2.l, 8, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v2, v2, s4, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_s_v2i8_s_v:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
+; GFX11-FAKE16-NEXT: global_load_u16 v1, v1, s[2:3]
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 8, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v2, v2, s4, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v1, s4, vcc_lo
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v2
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(4) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
@@ -501,26 +589,48 @@ define amdgpu_ps void @insertelement_s_v2i8_v_v(ptr addrspace(4) inreg %ptr, i8
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_s_v2i8_v_v:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
-; GFX11-NEXT: v_mov_b32_e32 v2, 0
-; GFX11-NEXT: global_load_u16 v2, v2, s[2:3]
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v3, 8, v2
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e32 v3, v3, v0, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-NEXT: v_dual_cndmask_b32 v0, v2, v0 :: v_dual_and_b32 v1, 0xff, v3
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_s_v2i8_v_v:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: global_load_u16 v2, v2, s[2:3]
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v3.l, 8, v2.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v3, v3, v0, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v0, vcc_lo
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v3.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_s_v2i8_v_v:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-FAKE16-NEXT: global_load_u16 v2, v2, s[2:3]
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 8, v2
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v0, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v0, v2, v0 :: v_dual_and_b32 v1, 0xff, v3
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(4) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
@@ -603,26 +713,47 @@ define amdgpu_ps void @insertelement_v_v2i8_s_v(ptr addrspace(1) %ptr, i8 inreg
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_v_v2i8_s_v:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: global_load_u16 v0, v[0:1], off
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_v_v2i8_s_v:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v2, v0, s2, vcc_lo
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_v_v2i8_s_v:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(1) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
@@ -702,26 +833,47 @@ define amdgpu_ps void @insertelement_v_v2i8_v_s(ptr addrspace(1) %ptr, i8 %val,
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_v_v2i8_v_s:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: global_load_u16 v0, v[0:1], off
-; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 0
-; GFX11-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_v_v2i8_v_s:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 0
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v0, v2, vcc_lo
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_v_v2i8_v_s:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 1
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e64 vcc_lo, s2, 0
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(1) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
@@ -801,26 +953,47 @@ define amdgpu_ps void @insertelement_v_v2i8_v_v(ptr addrspace(1) %ptr, i8 %val,
; GFX10-NEXT: global_store_short v[0:1], v2, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_v_v2i8_v_v:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: global_load_u16 v0, v[0:1], off
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v3
-; GFX11-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-NEXT: v_or_b32_e32 v2, v2, v3
-; GFX11-NEXT: global_store_b16 v[0:1], v2, off
-; GFX11-NEXT: s_endpgm
+; GFX11-TRUE16-LABEL: insertelement_v_v2i8_v_v:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v3
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v0, v2, vcc_lo
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
+; GFX11-TRUE16-NEXT: s_endpgm
+;
+; GFX11-FAKE16-LABEL: insertelement_v_v2i8_v_v:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v3
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
+; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
+; GFX11-FAKE16-NEXT: s_endpgm
%vec = load <2 x i8>, ptr addrspace(1) %ptr
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
store <2 x i8> %insert, ptr addrspace(1) null
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
index 6738c5e224267..6901d2240daed 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
@@ -6533,7 +6533,7 @@ define void @insert_very_small_from_very_large(<32 x i16> %L3, ptr %ptr) {
; GFX11-LABEL: insert_very_small_from_very_large:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_lshrrev_b16 v0.l, 1, v0.l
+; GFX11-NEXT: v_lshrrev_b32_e32 v0, 1, v0
; GFX11-NEXT: v_and_b16 v0.l, v0.l, 1
; GFX11-NEXT: v_lshlrev_b16 v0.l, 1, v0.l
; GFX11-NEXT: v_and_b32_e32 v0, 3, v0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir
index 30b716ad77311..25a62320299ba 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir
@@ -1929,10 +1929,17 @@ body: |
; GFX11-NEXT: {{ $}}
; GFX11-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
- ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<2 x s16>)
- ; GFX11-NEXT: [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<2 x s16>)
- ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
- ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+ ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+ ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
+ ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
; GFX11-NEXT: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT1]]
; GFX11-NEXT: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32)
; GFX11-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]]
@@ -1940,13 +1947,13 @@ body: |
; GFX11-NEXT: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT]], [[FMUL]]
; GFX11-NEXT: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA1]], [[FPEXT]]
; GFX11-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FMA2]], [[INT]]
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C]]
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C1]]
; GFX11-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[AND]], [[FMA1]]
; GFX11-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32)
- ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[UV2]](s16), [[UV]](s16)
- ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
- ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
+ ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC2]](s16), [[TRUNC]](s16)
+ ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16)
; GFX11-NEXT: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT3]]
; GFX11-NEXT: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32)
; GFX11-NEXT: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]]
@@ -1954,10 +1961,10 @@ body: |
; GFX11-NEXT: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FMA3]], [[INT2]], [[FMUL2]]
; GFX11-NEXT: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA4]], [[FPEXT2]]
; GFX11-NEXT: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FMA5]], [[INT2]]
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C]]
+ ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C1]]
; GFX11-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[AND1]], [[FMA4]]
; GFX11-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32)
- ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[UV3]](s16), [[UV1]](s16)
+ ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC3]](s16), [[TRUNC1]](s16)
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[INT1]](s16), [[INT3]](s16)
; GFX11-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
@@ -2660,10 +2667,27 @@ body: |
; GFX11-NEXT: {{ $}}
; GFX11-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
- ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
- ; GFX11-NEXT: [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
- ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
- ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV4]](s16)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+ ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+ ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+ ; GFX11-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+ ; GFX11-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+ ; GFX11-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
+ ; GFX11-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+ ; GFX11-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
+ ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16)
; GFX11-NEXT: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT1]]
; GFX11-NEXT: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32)
; GFX11-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]]
@@ -2671,13 +2695,13 @@ body: |
; GFX11-NEXT: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT]], [[FMUL]]
; GFX11-NEXT: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA1]], [[FPEXT]]
; GFX11-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FMA2]], [[INT]]
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C]]
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C1]]
; GFX11-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[AND]], [[FMA1]]
; GFX11-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32)
- ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[UV4]](s16), [[UV]](s16)
- ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
- ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV5]](s16)
+ ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC4]](s16), [[TRUNC]](s16)
+ ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16)
; GFX11-NEXT: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT3]]
; GFX11-NEXT: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32)
; GFX11-NEXT: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]]
@@ -2685,12 +2709,12 @@ body: |
; GFX11-NEXT: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FMA3]], [[INT2]], [[FMUL2]]
; GFX11-NEXT: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA4]], [[FPEXT2]]
; GFX11-NEXT: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FMA5]], [[INT2]]
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C]]
+ ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C1]]
; GFX11-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[AND1]], [[FMA4]]
; GFX11-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32)
- ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[UV5]](s16), [[UV1]](s16)
- ; GFX11-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
- ; GFX11-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[UV6]](s16)
+ ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC5]](s16), [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
+ ; GFX11-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16)
; GFX11-NEXT: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT5]]
; GFX11-NEXT: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32)
; GFX11-NEXT: [[FMUL4:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]]
@@ -2698,12 +2722,12 @@ body: |
; GFX11-NEXT: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FMA6]], [[INT4]], [[FMUL4]]
; GFX11-NEXT: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA7]], [[FPEXT4]]
; GFX11-NEXT: [[FMUL5:%[0-9]+]]:_(s32) = G_FMUL [[FMA8]], [[INT4]]
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[FMUL5]], [[C]]
+ ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[FMUL5]], [[C1]]
; GFX11-NEXT: [[FADD2:%[0-9]+]]:_(s32) = G_FADD [[AND2]], [[FMA7]]
; GFX11-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD2]](s32)
- ; GFX11-NEXT: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[UV6]](s16), [[UV2]](s16)
- ; GFX11-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
- ; GFX11-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[UV7]](s16)
+ ; GFX11-NEXT: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC6]](s16), [[TRUNC2]](s16)
+ ; GFX11-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16)
+ ; GFX11-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16)
; GFX11-NEXT: [[FNEG3:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT7]]
; GFX11-NEXT: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT7]](s32)
; GFX11-NEXT: [[FMUL6:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT6]], [[INT6]]
@@ -2711,10 +2735,10 @@ body: |
; GFX11-NEXT: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FMA9]], [[INT6]], [[FMUL6]]
; GFX11-NEXT: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FNEG3]], [[FMA10]], [[FPEXT6]]
; GFX11-NEXT: [[FMUL7:%[0-9]+]]:_(s32) = G_FMUL [[FMA11]], [[INT6]]
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[FMUL7]], [[C]]
+ ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[FMUL7]], [[C1]]
; GFX11-NEXT: [[FADD3:%[0-9]+]]:_(s32) = G_FADD [[AND3]], [[FMA10]]
; GFX11-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD3]](s32)
- ; GFX11-NEXT: [[INT7:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC3]](s16), [[UV7]](s16), [[UV3]](s16)
+ ; GFX11-NEXT: [[INT7:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC3]](s16), [[TRUNC7]](s16), [[TRUNC3]](s16)
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[INT1]](s16), [[INT3]](s16)
; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[INT5]](s16), [[INT7]](s16)
; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
index a5a76bb18e557..530662cabd94c 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
@@ -2,8 +2,9 @@
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=bonaire -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=CI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9 %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9 %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX10 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
---
name: test_load_constant_s1_align1
@@ -37,6 +38,24 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s1_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s1_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[AND]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s1) = G_LOAD %0 :: (load (s1), align 1, addrspace 4)
%2:_(s32) = G_ZEXT %1
@@ -75,6 +94,24 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s2_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s2_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[AND]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s2) = G_LOAD %0 :: (load (s2), align 1, addrspace 4)
%2:_(s32) = G_ZEXT %1
@@ -107,6 +144,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s8_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s8_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s8) = G_LOAD %0 :: (load (s8), align 4, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -139,6 +190,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s8_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s8_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s8) = G_LOAD %0 :: (load (s8), align 1, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -171,6 +236,28 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_s16_align4
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_s16_align4
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 4, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -203,6 +290,28 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s16_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 2, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -253,6 +362,32 @@ body: |
; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s16_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s16_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 1, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -285,6 +420,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32), align 4, addrspace 4)
$vgpr0 = COPY %1
@@ -334,6 +483,32 @@ body: |
; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32), align 2, addrspace 4)
$vgpr0 = COPY %1
@@ -413,6 +588,52 @@ body: |
; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32), align 1, addrspace 4)
$vgpr0 = COPY %1
@@ -444,6 +665,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s24_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), align 8, addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s24_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), align 8, addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 8, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -476,6 +711,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s24_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s24_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 4, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -526,6 +775,32 @@ body: |
; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s24_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 2, align 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s24_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 2, align 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 2, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -594,6 +869,44 @@ body: |
; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
; GFX9-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s24_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s24_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR1]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 1, addrspace 4)
%2:_(s32) = G_ANYEXT %1
@@ -632,6 +945,24 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s48_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s48_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s48) = G_LOAD %0 :: (load (s48), align 8, addrspace 4)
%2:_(s64) = G_ZEXT %1
@@ -664,6 +995,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s64_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s64_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 8, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -695,6 +1040,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s64_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s64_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -780,6 +1139,56 @@ body: |
; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s64_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s64_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 2, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -919,6 +1328,92 @@ body: |
; GFX9-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s64_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s64_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 1, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -953,6 +1448,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s96_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s96_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -987,6 +1498,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s96_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s96_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -1021,6 +1548,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s96_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s96_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -1118,6 +1661,64 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s96_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s96_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 2, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -1293,6 +1894,116 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s96_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s96_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX11-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 1, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -1342,6 +2053,32 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; GFX9-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s160_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s32) from unknown-address + 16, addrspace 4)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX10-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s160_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s32) from unknown-address + 16, addrspace 4)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX11-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s160) = G_LOAD %0 :: (load (s160), align 4, addrspace 4)
S_NOP 0, implicit %1
@@ -1400,6 +2137,38 @@ body: |
; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s224_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 4)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s224_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 4)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 4)
%2:_(s256) = G_IMPLICIT_DEF
@@ -1437,6 +2206,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s128_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s128_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -1471,6 +2256,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s128_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s128_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -1691,6 +2492,146 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s128_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s128_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX11-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX11-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX11-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX11-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX11-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 1, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -1725,6 +2666,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s256_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s256_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s256) = G_LOAD %0 :: (load (s256), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -1756,6 +2713,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p1_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p1_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p1) = G_LOAD %0 :: (load (p1), align 8, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -1787,6 +2758,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p1_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p1_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p1) = G_LOAD %0 :: (load (p1), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -1929,6 +2914,94 @@ body: |
; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p1_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p1_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p1) = G_LOAD %0 :: (load (p1), align 1, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -1960,6 +3033,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p4) :: (load (p3), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p3_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p4) :: (load (p3), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p3_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p4) :: (load (p3), addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](p3)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p3) = G_LOAD %0 :: (load (p3), align 4, addrspace 4)
$vgpr0 = COPY %1
@@ -1991,6 +3078,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p4_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p4_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 8, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -2022,6 +3123,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p4_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p4_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -2110,6 +3225,58 @@ body: |
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p4_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p4_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX11-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 2, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -2252,6 +3419,94 @@ body: |
; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p4_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p4_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 1, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -2283,6 +3538,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p4) :: (load (p5), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p5_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p4) :: (load (p5), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p5_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p4) :: (load (p5), addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](p5)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p5) = G_LOAD %0 :: (load (p5), align 4, addrspace 4)
$vgpr0 = COPY %1
@@ -2335,6 +3604,34 @@ body: |
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
; GFX9-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p5_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
+ ; GFX10-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p5_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
+ ; GFX11-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p5) = G_LOAD %0 :: (load (p5), align 2, addrspace 4)
$vgpr0 = COPY %1
@@ -2417,6 +3714,54 @@ body: |
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
; GFX9-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_p5_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
+ ; GFX10-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_p5_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
+ ; GFX11-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(p5) = G_LOAD %0 :: (load (p5), align 1, addrspace 4)
$vgpr0 = COPY %1
@@ -2448,6 +3793,28 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s8_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v2s8_align4
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v2s8_align4
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 4, addrspace 4)
%2:_(s16) = G_BITCAST %1
@@ -2481,6 +3848,28 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s8_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v2s8_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v2s8_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 2, addrspace 4)
%2:_(s16) = G_BITCAST %1
@@ -2532,6 +3921,32 @@ body: |
; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s8_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s8_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 1, addrspace 4)
%2:_(s16) = G_BITCAST %1
@@ -2636,6 +4051,66 @@ body: |
; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s8_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s8_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 4, addrspace 4)
%2:_(s24) = G_BITCAST %1
@@ -2770,6 +4245,86 @@ body: |
; GFX9-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; GFX9-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s8_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s8_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR4]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 4)
%2:_(s24) = G_BITCAST %1
@@ -2803,6 +4358,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s8_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s8_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 4, addrspace 4)
%2:_(s32) = G_BITCAST %1
@@ -2853,6 +4422,32 @@ body: |
; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s8_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s8_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 2, addrspace 4)
%2:_(s32) = G_BITCAST %1
@@ -2933,6 +4528,52 @@ body: |
; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s8_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s8_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 1, addrspace 4)
%2:_(s32) = G_BITCAST %1
@@ -2965,6 +4606,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v8s8_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v8s8_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<8 x s8>) = G_LOAD %0 :: (load (<8 x s8>), align 8, addrspace 4)
%2:_(<2 x s32>) = G_BITCAST %1
@@ -3096,6 +4751,118 @@ body: |
; GFX9-NEXT: [[TRUNC19:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
; GFX9-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v16s8_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+ ; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[TRUNC6]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC7]](s16), [[TRUNC8]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC9:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+ ; GFX10-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX10-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX10-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC14:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS2]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+ ; GFX10-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX10-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX10-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC15]](s16), [[TRUNC16]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC17]](s16), [[TRUNC18]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC19:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
+ ; GFX10-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v16s8_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+ ; GFX11-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX11-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX11-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[TRUNC6]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC7]](s16), [[TRUNC8]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC9:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+ ; GFX11-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX11-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX11-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC14:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS2]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+ ; GFX11-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX11-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX11-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC15]](s16), [[TRUNC16]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC17]](s16), [[TRUNC18]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC19:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
+ ; GFX11-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<16 x s8>) = G_LOAD %0 :: (load (<16 x s8>), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -3310,6 +5077,206 @@ body: |
; GFX9-NEXT: [[TRUNC39:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS7]](<4 x s16>)
; GFX9-NEXT: [[CONCAT_VECTORS8:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>), [[TRUNC24]](<4 x s8>), [[TRUNC29]](<4 x s8>), [[TRUNC34]](<4 x s8>), [[TRUNC39]](<4 x s8>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v32s8_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+ ; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[TRUNC6]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC7]](s16), [[TRUNC8]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC9:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+ ; GFX10-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX10-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX10-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC14:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS2]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+ ; GFX10-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX10-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX10-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC15]](s16), [[TRUNC16]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC17]](s16), [[TRUNC18]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC19:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+ ; GFX10-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX10-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX10-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC24:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS4]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
+ ; GFX10-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX10-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX10-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC25]](s16), [[TRUNC26]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC27]](s16), [[TRUNC28]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS5:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC29:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS5]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
+ ; GFX10-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX10-NEXT: [[TRUNC32:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX10-NEXT: [[TRUNC33:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC32]](s16), [[TRUNC33]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS6:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC34:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS6]](<4 x s16>)
+ ; GFX10-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C]](s32)
+ ; GFX10-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C1]](s32)
+ ; GFX10-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C2]](s32)
+ ; GFX10-NEXT: [[TRUNC35:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
+ ; GFX10-NEXT: [[TRUNC36:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX10-NEXT: [[TRUNC37:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX10-NEXT: [[TRUNC38:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC35]](s16), [[TRUNC36]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC37]](s16), [[TRUNC38]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS7:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX10-NEXT: [[TRUNC39:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS7]](<4 x s16>)
+ ; GFX10-NEXT: [[CONCAT_VECTORS8:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>), [[TRUNC24]](<4 x s8>), [[TRUNC29]](<4 x s8>), [[TRUNC34]](<4 x s8>), [[TRUNC39]](<4 x s8>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v32s8_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+ ; GFX11-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX11-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX11-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[TRUNC6]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC7]](s16), [[TRUNC8]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC9:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+ ; GFX11-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX11-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX11-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC14:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS2]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+ ; GFX11-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX11-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX11-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC15]](s16), [[TRUNC16]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC17]](s16), [[TRUNC18]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC19:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+ ; GFX11-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX11-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX11-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC24:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS4]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
+ ; GFX11-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX11-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX11-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC25]](s16), [[TRUNC26]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC27]](s16), [[TRUNC28]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS5:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC29:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS5]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
+ ; GFX11-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX11-NEXT: [[TRUNC32:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX11-NEXT: [[TRUNC33:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC32]](s16), [[TRUNC33]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS6:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC34:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS6]](<4 x s16>)
+ ; GFX11-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C]](s32)
+ ; GFX11-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C1]](s32)
+ ; GFX11-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C2]](s32)
+ ; GFX11-NEXT: [[TRUNC35:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
+ ; GFX11-NEXT: [[TRUNC36:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX11-NEXT: [[TRUNC37:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX11-NEXT: [[TRUNC38:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC35]](s16), [[TRUNC36]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC37]](s16), [[TRUNC38]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS7:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX11-NEXT: [[TRUNC39:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS7]](<4 x s16>)
+ ; GFX11-NEXT: [[CONCAT_VECTORS8:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>), [[TRUNC24]](<4 x s8>), [[TRUNC29]](<4 x s8>), [[TRUNC34]](<4 x s8>), [[TRUNC39]](<4 x s8>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<32 x s8>) = G_LOAD %0 :: (load (<32 x s8>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -3342,6 +5309,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p4) :: (load (<2 x s16>), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p4) :: (load (<2 x s16>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s16_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p4) :: (load (<2 x s16>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 4, addrspace 4)
$vgpr0 = COPY %1
@@ -3399,6 +5380,43 @@ body: |
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s16_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v2s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v2s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 4)
$vgpr0 = COPY %1
@@ -3486,6 +5504,52 @@ body: |
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s16_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s16_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 1, addrspace 4)
$vgpr0 = COPY %1
@@ -3535,6 +5599,28 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s16_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s16_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 4)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -3626,6 +5712,66 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, align 4, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v3s16_align4
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, align 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v3s16_align4
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, align 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 4, addrspace 4)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -3717,6 +5863,66 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s16_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v3s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v3s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 2, addrspace 4)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -3850,6 +6056,76 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s16_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s16_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 1, addrspace 4)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -3883,6 +6159,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s16_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s16_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 8, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -3914,6 +6204,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s16_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -4005,6 +6309,71 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s16_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v4s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v4s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 2, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -4150,6 +6519,88 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s16_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s16_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 1, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -4184,6 +6635,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v8s16_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v8s16_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4215,6 +6682,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 8, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -4246,6 +6727,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -4319,6 +6814,48 @@ body: |
; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 2, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -4446,6 +6983,84 @@ body: |
; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 1, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -4477,6 +7092,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -4510,6 +7139,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -4541,6 +7184,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4572,6 +7229,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4603,6 +7274,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4634,6 +7319,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v8s32_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v8s32_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = G_LOAD %0 :: (load (<8 x s32>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -4665,6 +7364,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v16s32_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v16s32_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<16 x s32>) = G_LOAD %0 :: (load (<16 x s32>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
@@ -4696,6 +7409,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s16>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v16s32_align32_extload_from_v16s16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s16>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v16s32_align32_extload_from_v16s16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s16>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<16 x s32>) = G_LOAD %0 :: (load (<16 x s16>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
@@ -4727,6 +7454,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s64_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s64_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4758,6 +7499,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s64_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s64_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4789,6 +7544,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s64_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s64_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4931,6 +7700,94 @@ body: |
; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s64_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s16) from unknown-address + 12, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s16) from unknown-address + 14, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s64_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s16) from unknown-address + 12, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s16) from unknown-address + 14, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 2, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -5175,6 +8032,162 @@ body: |
; GFX9-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s64_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s64_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX11-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX11-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX11-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX11-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX11-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX11-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX11-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX11-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX11-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 1, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -5206,6 +8219,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s64_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s64_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 4)
%2:_(<4 x s64>) = G_IMPLICIT_DEF
@@ -5257,6 +8284,32 @@ body: |
; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s64_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p4) :: (load (s64) from unknown-address + 16, addrspace 4)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s64_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p4) :: (load (s64) from unknown-address + 16, addrspace 4)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 8, addrspace 4)
%2:_(<4 x s64>) = G_IMPLICIT_DEF
@@ -5608,6 +8661,232 @@ body: |
; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v3s64_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX10-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 4)
+ ; GFX10-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX10-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 4)
+ ; GFX10-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX10-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX10-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX10-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 4)
+ ; GFX10-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX10-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 4)
+ ; GFX10-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX10-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX10-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX10-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v3s64_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX11-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX11-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX11-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX11-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX11-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX11-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX11-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX11-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX11-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX11-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 4)
+ ; GFX11-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX11-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 4)
+ ; GFX11-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX11-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX11-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX11-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 4)
+ ; GFX11-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX11-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 4)
+ ; GFX11-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX11-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX11-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX11-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 1, addrspace 4)
%2:_(<4 x s64>) = G_IMPLICIT_DEF
@@ -5641,6 +8920,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s64_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s64_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -5672,6 +8965,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s64_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), align 8, addrspace 4)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s64_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), align 8, addrspace 4)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -6120,6 +9427,298 @@ body: |
; GFX9-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s64_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX10-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 4)
+ ; GFX10-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX10-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 4)
+ ; GFX10-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX10-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX10-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX10-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 4)
+ ; GFX10-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX10-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 4)
+ ; GFX10-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX10-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX10-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX10-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX10-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
+ ; GFX10-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C8]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p4) :: (load (s8) from unknown-address + 24, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p4) :: (load (s8) from unknown-address + 25, addrspace 4)
+ ; GFX10-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
+ ; GFX10-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p4) :: (load (s8) from unknown-address + 26, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p4) :: (load (s8) from unknown-address + 27, addrspace 4)
+ ; GFX10-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
+ ; GFX10-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
+ ; GFX10-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
+ ; GFX10-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p4) :: (load (s8) from unknown-address + 28, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p4) :: (load (s8) from unknown-address + 29, addrspace 4)
+ ; GFX10-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
+ ; GFX10-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p4) :: (load (s8) from unknown-address + 30, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p4) :: (load (s8) from unknown-address + 31, addrspace 4)
+ ; GFX10-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
+ ; GFX10-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
+ ; GFX10-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
+ ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
+ ; GFX10-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s64_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX11-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX11-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX11-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX11-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX11-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX11-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX11-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX11-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX11-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX11-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 4)
+ ; GFX11-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX11-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 4)
+ ; GFX11-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX11-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX11-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX11-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 4)
+ ; GFX11-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX11-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 4)
+ ; GFX11-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX11-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX11-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX11-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX11-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
+ ; GFX11-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C8]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p4) :: (load (s8) from unknown-address + 24, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p4) :: (load (s8) from unknown-address + 25, addrspace 4)
+ ; GFX11-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
+ ; GFX11-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p4) :: (load (s8) from unknown-address + 26, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p4) :: (load (s8) from unknown-address + 27, addrspace 4)
+ ; GFX11-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
+ ; GFX11-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
+ ; GFX11-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
+ ; GFX11-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p4) :: (load (s8) from unknown-address + 28, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p4) :: (load (s8) from unknown-address + 29, addrspace 4)
+ ; GFX11-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
+ ; GFX11-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p4) :: (load (s8) from unknown-address + 30, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p4) :: (load (s8) from unknown-address + 31, addrspace 4)
+ ; GFX11-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
+ ; GFX11-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
+ ; GFX11-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
+ ; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
+ ; GFX11-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 1, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -6154,6 +9753,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2s128_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2s128_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s128>) = G_LOAD %0 :: (load (<2 x s128>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -6188,6 +9803,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2p1_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2p1_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -6222,6 +9853,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2p1_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2p1_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -6256,6 +9903,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2p1_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2p1_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -6476,6 +10139,146 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2p1_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2p1_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
+ ; GFX11-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
+ ; GFX11-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX11-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
+ ; GFX11-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX11-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 1, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -6510,6 +10313,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2p3_align8
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2p3_align8
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 8, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -6544,6 +10363,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2p3_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2p3_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -6674,6 +10509,86 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v2p3_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v2p3_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 1, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -6705,6 +10620,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s32_from_1_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_ext_load_constant_s32_from_1_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s8), align 4, addrspace 4)
$vgpr0 = COPY %1
@@ -6736,6 +10665,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s32_from_2_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-LABEL: name: test_ext_load_constant_s32_from_2_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s16), align 4, addrspace 4)
$vgpr0 = COPY %1
@@ -6771,6 +10714,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s64_from_1_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-LABEL: name: test_ext_load_constant_s64_from_1_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -6805,6 +10764,30 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s64_from_2_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_ext_load_constant_s64_from_2_align4
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_ext_load_constant_s64_from_2_align4
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -6839,6 +10822,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s64_from_4_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-LABEL: name: test_ext_load_constant_s64_from_4_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s32), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -6882,6 +10881,28 @@ body: |
; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s128_from_4_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX10-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX11-LABEL: name: test_ext_load_constant_s128_from_4_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX11-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s32), align 4, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -6916,6 +10937,30 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s64_from_2_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_ext_load_constant_s64_from_2_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_ext_load_constant_s64_from_2_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -6950,6 +10995,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-LABEL: name: test_ext_load_constant_s64_from_1_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-LABEL: name: test_ext_load_constant_s64_from_1_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
+ ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 4)
$vgpr0_vgpr1 = COPY %1
@@ -7077,6 +11138,84 @@ body: |
; GFX9-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v2s32_from_4_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v2s32_from_4_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -7150,6 +11289,48 @@ body: |
; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v2s32_from_4_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v2s32_from_4_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 2, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -7181,6 +11362,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 1)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v2s32_from_4_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v2s32_from_4_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -7212,6 +11407,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v3s32_from_6_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v3s32_from_6_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -7243,6 +11452,20 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v4s32_from_8_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v4s32_from_8_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -7562,6 +11785,212 @@ body: |
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align1
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX10-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
+ ; GFX10-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
+ ; GFX10-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
+ ; GFX10-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX10-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
+ ; GFX10-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX10-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
+ ; GFX10-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
+ ; GFX10-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
+ ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align1
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX11-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX11-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX11-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX11-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX11-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
+ ; GFX11-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX11-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
+ ; GFX11-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
+ ; GFX11-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX11-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
+ ; GFX11-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX11-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
+ ; GFX11-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
+ ; GFX11-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
+ ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 1, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -7734,6 +12163,112 @@ body: |
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align2
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s16) from unknown-address + 16, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p4) :: (load (s16) from unknown-address + 18, addrspace 1)
+ ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s64)
+ ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s16) from unknown-address + 20, addrspace 1)
+ ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s16) from unknown-address + 22, addrspace 1)
+ ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
+ ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align2
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s16) from unknown-address + 16, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p4) :: (load (s16) from unknown-address + 18, addrspace 1)
+ ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX11-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s64)
+ ; GFX11-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s16) from unknown-address + 20, addrspace 1)
+ ; GFX11-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s16) from unknown-address + 22, addrspace 1)
+ ; GFX11-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
+ ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 2, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -7792,6 +12327,36 @@ body: |
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align4
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align4
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 4, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -7850,6 +12415,36 @@ body: |
; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align16
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align16
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 16, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -7887,6 +12482,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_s512_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_s512_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s512) = G_LOAD %0 :: (load (s512), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
@@ -7921,6 +12532,22 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
+ ;
+ ; GFX10-LABEL: name: test_load_constant_v4s128_align32
+ ; GFX10: liveins: $vgpr0_vgpr1
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
+ ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
+ ;
+ ; GFX11-LABEL: name: test_load_constant_v4s128_align32
+ ; GFX11: liveins: $vgpr0_vgpr1
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<4 x s128>) = G_LOAD %0 :: (load (<4 x s128>), align 32, addrspace 4)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
index f1c9a77987f30..397dcc4021919 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
@@ -2,14 +2,16 @@
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=CI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9PLUS %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9PLUS %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11PLUS %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX12 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX10PLUS %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11PLUS,GFX11PLUS-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11PLUS,GFX11PLUS-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX12,GFX12-FAKE16 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=CI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX9PLUS %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX9PLUS %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX10PLUS %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX11PLUS %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX12 %s
@@ -46,6 +48,15 @@ body: |
; GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s1_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -73,6 +84,15 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s1_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -129,6 +149,15 @@ body: |
; GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s2_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s2_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -156,6 +185,15 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s2_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s2_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -206,6 +244,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s8_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s8_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -227,6 +272,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s8_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -273,6 +325,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s8_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s8_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -294,6 +353,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s8_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -340,19 +406,42 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_s16_align4
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_s16_align4
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX10PLUS-LABEL: name: test_load_flat_s16_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align4
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_s16_align4
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_s16_align4
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_s16_align4
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_s16_align4
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -361,6 +450,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s16_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -407,19 +503,42 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_s16_align2
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_s16_align2
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX10PLUS-LABEL: name: test_load_flat_s16_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align2
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_s16_align2
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_s16_align2
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -428,6 +547,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s16_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -486,19 +612,42 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_s16_align1
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_s16_align1
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX10PLUS-LABEL: name: test_load_flat_s16_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align1
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_s16_align1
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_s16_align1
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_s16_align1
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_s16_align1
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -513,6 +662,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s16_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -571,6 +733,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s32_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -592,6 +761,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -649,6 +825,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s32_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -676,6 +859,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -765,6 +961,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s32_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -802,6 +1005,29 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -913,6 +1139,15 @@ body: |
; GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s48_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s48_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -940,6 +1175,15 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s48_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s48_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -998,6 +1242,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s64_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1019,6 +1270,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1072,6 +1330,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s64_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1093,6 +1358,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1166,6 +1438,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s64_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1205,6 +1484,31 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1350,6 +1654,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s64_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1407,6 +1718,49 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1541,6 +1895,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s96_align16
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1565,6 +1927,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align16
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1629,6 +1999,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s96_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1653,6 +2031,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1717,6 +2103,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s96_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1741,6 +2135,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1833,6 +2235,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s96_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 2)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1878,6 +2288,35 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2064,6 +2503,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s96_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 1)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2135,6 +2582,61 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2310,6 +2812,19 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; GFX9PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s160_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX10PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s160_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2349,6 +2864,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s160_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s160_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2459,6 +2987,22 @@ body: |
; GFX9PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s224_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
+ ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s224_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2507,6 +3051,22 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s224_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s224_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2596,6 +3156,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s128_align16
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s128_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2620,6 +3188,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s128_align16
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s128_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2690,6 +3266,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s128_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s128_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2714,6 +3298,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s128_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s128_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2888,6 +3480,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s128_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 1)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s128_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2974,6 +3574,76 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s128_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s128_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3196,6 +3866,18 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s256_align32
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s256_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3232,6 +3914,18 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s256_align32
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s256_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3295,6 +3989,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p1_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p1_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3316,6 +4017,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p1_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p1_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3369,6 +4077,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p1_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3390,6 +4105,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p1_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3499,6 +4221,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p1_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3557,6 +4286,50 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p1_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3676,6 +4449,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p3_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p3_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3697,6 +4477,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p3_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p3_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3750,6 +4537,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p4_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3771,6 +4565,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3824,6 +4625,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p4_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3845,6 +4653,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3918,6 +4733,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p4_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3958,6 +4780,32 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4105,6 +4953,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p4_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4163,6 +5018,50 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4282,6 +5181,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p5_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p5_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4303,6 +5209,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p5_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p5_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4362,6 +5275,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p5_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p5_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4390,6 +5310,20 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p5_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p5_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4483,6 +5417,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_p5_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_p5_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4521,6 +5462,30 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p5_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p5_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4600,19 +5565,42 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_v2s8_align4
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_v2s8_align4
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s8_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align4
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v2s8_align4
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v2s8_align4
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v2s8_align4
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v2s8_align4
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -4621,6 +5609,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s8_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4668,19 +5663,42 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_v2s8_align2
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_v2s8_align2
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s8_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align2
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v2s8_align2
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v2s8_align2
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v2s8_align2
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v2s8_align2
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -4689,6 +5707,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s8_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4748,19 +5773,42 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_v2s8_align1
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_v2s8_align1
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s8_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align1
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v2s8_align1
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v2s8_align1
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v2s8_align1
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v2s8_align1
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -4775,6 +5823,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s8_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4905,6 +5966,36 @@ body: |
; GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s8_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s8_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4995,6 +6086,36 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s8_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5184,6 +6305,41 @@ body: |
; GFX9PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[OR3]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s8_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
+ ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s8_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5294,6 +6450,46 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR4]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s8_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5407,6 +6603,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s8_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s8_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5428,6 +6631,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s8_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5486,6 +6696,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s8_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s8_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5513,6 +6730,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s8_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s8_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5603,6 +6833,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s8_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s8_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5640,6 +6877,29 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s8_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5726,6 +6986,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v8s8_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v8s8_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5747,6 +7014,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v8s8_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v8s8_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5813,6 +7087,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v16s8_align16
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v16s8_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5834,6 +7115,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v16s8_align16
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v16s8_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5928,6 +7216,17 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v32s8_align32
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v32s8_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5961,6 +7260,17 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v32s8_align32
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v32s8_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6016,6 +7326,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s16_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s16_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -6037,6 +7354,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s16_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6102,6 +7426,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s16_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s16_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -6129,6 +7460,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s16_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6226,6 +7570,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s16_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s16_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -6263,6 +7614,29 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s16_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6384,6 +7758,17 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s16_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -6417,6 +7802,17 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6526,47 +7922,104 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_v3s16_align4
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX11PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX11PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
- ; GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX12-LABEL: name: test_load_flat_v3s16_align4
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align4
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11PLUS-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; GFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s16_align4
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11PLUS-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v3s16_align4
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v3s16_align4
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v3s16_align4
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -6589,6 +8042,27 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6722,47 +8196,104 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_v3s16_align2
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX11PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX11PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX12-LABEL: name: test_load_flat_v3s16_align2
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align2
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11PLUS-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; GFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s16_align2
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11PLUS-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v3s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v3s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v3s16_align2
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -6785,6 +8316,27 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6946,47 +8498,104 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11PLUS-LABEL: name: test_load_flat_v3s16_align1
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
- ; GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX11PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX11PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
- ; GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX12-LABEL: name: test_load_flat_v3s16_align1
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
+ ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
+ ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align1
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
+ ; GFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11PLUS-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
+ ; GFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s16_align1
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11PLUS-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v3s16_align1
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v3s16_align1
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v3s16_align1
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -7023,6 +8632,41 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7134,6 +8778,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7155,6 +8806,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7208,6 +8866,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7229,6 +8894,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7316,6 +8988,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7353,6 +9032,29 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7508,6 +9210,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7563,6 +9272,47 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7699,6 +9449,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v8s16_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v8s16_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7723,6 +9481,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v8s16_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v8s16_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7778,6 +9544,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s32_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s32_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7799,6 +9572,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s32_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s32_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7852,6 +9632,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s32_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7873,6 +9660,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s32_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7927,6 +9721,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s32_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s32_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7948,6 +9749,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s32_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s32_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8007,6 +9815,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s32_align16
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s32_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8028,6 +9843,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s32_align16
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s32_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8089,6 +9911,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s32_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8110,6 +9939,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s32_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8175,6 +10011,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s32_align16
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s32_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8196,6 +10039,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s32_align16
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s32_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8261,6 +10111,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s32_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s32_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8282,6 +10139,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s32_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s32_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8347,6 +10211,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s32_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8368,6 +10239,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s32_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8461,6 +10339,17 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v8s32_align32
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v8s32_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8494,6 +10383,17 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v8s32_align32
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v8s32_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8649,6 +10549,23 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v16s32_align32
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p0) :: (load (<4 x s32>) from unknown-address + 32, align 32)
+ ; GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 48
+ ; GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p0) :: (load (<4 x s32>) from unknown-address + 48)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v16s32_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8700,6 +10617,23 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v16s32_align32
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p0) :: (load (<4 x s32>) from unknown-address + 32, align 32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 48
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p0) :: (load (<4 x s32>) from unknown-address + 48)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v16s32_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8787,6 +10721,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align16
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8808,6 +10749,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align16
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8875,6 +10823,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8896,6 +10851,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8963,6 +10925,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8984,6 +10953,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9087,6 +11063,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 2)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9145,6 +11128,50 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s16) from unknown-address + 12)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s16) from unknown-address + 14)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9390,6 +11417,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9482,6 +11516,84 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9711,6 +11823,19 @@ body: |
; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s64_align32
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
+ ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s64_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9750,6 +11875,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s64_align32
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s64_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9851,6 +11989,19 @@ body: |
; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s64_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9890,6 +12041,19 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s64_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10143,6 +12307,19 @@ body: |
; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v3s64_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 1)
+ ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10282,6 +12459,119 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s64_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10591,6 +12881,17 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s64_align32
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s64_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10624,6 +12925,17 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s64_align32
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s64_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10727,6 +13039,17 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s64_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 8)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10760,6 +13083,17 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s64_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11063,6 +13397,17 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v4s64_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 1)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11232,6 +13577,153 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s64>), [[BUILD_VECTOR1]](<2 x s64>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s64_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p0) :: (load (s8) from unknown-address + 24)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p0) :: (load (s8) from unknown-address + 25)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p0) :: (load (s8) from unknown-address + 26)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p0) :: (load (s8) from unknown-address + 27)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p0) :: (load (s8) from unknown-address + 28)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p0) :: (load (s8) from unknown-address + 29)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p0) :: (load (s8) from unknown-address + 30)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p0) :: (load (s8) from unknown-address + 31)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR20]](s64), [[OR27]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s64>), [[BUILD_VECTOR1]](<2 x s64>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11608,6 +14100,18 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2s128_align32
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s128_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11644,6 +14148,18 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s128_align32
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s128_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11722,6 +14238,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align16
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11746,6 +14270,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align16
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11816,6 +14348,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11840,6 +14380,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11910,6 +14458,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11934,6 +14490,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12108,6 +14672,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 1)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12194,6 +14766,76 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12376,6 +15018,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2p3_align8
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p3_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12400,6 +15050,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p3_align8
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p3_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12458,6 +15116,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2p3_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p3_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12482,6 +15148,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p3_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p3_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12596,6 +15270,14 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_v2p3_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 1)
+ ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p3_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12652,6 +15334,46 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p3_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p3_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12763,6 +15485,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12784,6 +15513,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12829,6 +15565,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12850,6 +15593,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12899,6 +15649,14 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12923,6 +15681,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12973,21 +15739,45 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX11PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX12-LABEL: name: test_ext_load_flat_s64_from_2_align4
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -12997,6 +15787,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -13047,6 +15845,14 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -13071,6 +15877,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -13130,6 +15944,17 @@ body: |
; GFX9PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
;
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX10PLUS-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -13163,6 +15988,17 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX10PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -13219,21 +16055,45 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX11PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX12-LABEL: name: test_ext_load_flat_s64_from_2_align2
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11PLUS-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX11PLUS-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -13243,6 +16103,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -13293,6 +16161,14 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -13317,6 +16193,14 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
+ ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -13365,6 +16249,13 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; GFX10PLUS-LABEL: name: test_load_flat_s32_align536870912
+ ; GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: {{ $}}
+ ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
+ ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align536870912
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -13386,6 +16277,13 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
+ ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align536870912
+ ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
+ ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
+ ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align536870912
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
index a15c602e94878..77d72f62b5e3a 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
@@ -12,12 +12,16 @@
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-MESA %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-HSA %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-MESA %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-HSA %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-MESA %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-HSA %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-MESA %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX10-HSA %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX10-MESA %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-HSA,GFX11-HSA-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-HSA,GFX11-HSA-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-MESA,GFX11-MESA-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-MESA,GFX11-MESA-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12-HSA,GFX12-HSA-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12-HSA,GFX12-HSA-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12-MESA,GFX-MESA-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12-MESA,GFX-MESA-FAKE16 %s
# ERR-NOT: remark
# ERR: remark: <unknown>:0:0: unable to legalize instruction: %{{[0-9]+}}:_(<2 x s32>) = G_LOAD %{{[0-9]+}}:_(p1) :: (load (<2 x s16>), align 1, addrspace 1) (in function: test_extload_global_v2s32_from_v2s16_align1)
@@ -86,6 +90,60 @@ body: |
; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX9-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s1_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s1_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s1_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s1_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s1_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s1_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s1) = G_LOAD %0 :: (load (s1), align 1, addrspace 1)
%2:_(s32) = G_ZEXT %1
@@ -151,6 +209,60 @@ body: |
; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; GFX9-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s2_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s2_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s2_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX11-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s2_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX11-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s2_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX12-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s2_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX12-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s2) = G_LOAD %0 :: (load (s2), align 1, addrspace 1)
%2:_(s32) = G_ZEXT %1
@@ -204,6 +316,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s8_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s8_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s8_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s8_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s8_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s8_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s8) = G_LOAD %0 :: (load (s8), align 4, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -257,6 +411,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s8_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s8_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s8_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s8_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s8_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s8_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s8) = G_LOAD %0 :: (load (s8), align 1, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -310,6 +506,80 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_s16_align4
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_s16_align4
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_s16_align4
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_s16_align4
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_s16_align4
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_s16_align4
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_s16_align4
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_s16_align4
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 4, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -363,6 +633,80 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_s16_align2
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_s16_align2
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_s16_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_s16_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_s16_align2
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_s16_align2
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_s16_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_s16_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 2, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -440,6 +784,82 @@ body: |
; GFX9-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_s16_align1
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_s16_align1
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_s16_align1
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_s16_align1
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 1, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -493,6 +913,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s32_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s32_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s32_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s32_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s32_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s32_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -569,6 +1031,66 @@ body: |
; GFX9-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s32_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s32_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s32_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s32_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s32_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s32_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32), align 2, addrspace 1)
$vgpr0 = COPY %1
@@ -685,6 +1207,96 @@ body: |
; GFX9-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s32_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s32_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s32_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s32_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s32_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s32_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s32), align 1, addrspace 1)
$vgpr0 = COPY %1
@@ -737,6 +1349,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s24_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s24_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s24_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s24_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s24_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s24_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 8, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -790,6 +1444,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s24_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s24_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s24_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s24_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s24_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s24_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 4, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -879,6 +1575,84 @@ body: |
; GFX9-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s24_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s24_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s24_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s24_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s24_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s24_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 2, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -992,6 +1766,102 @@ body: |
; GFX9-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
; GFX9-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s24_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s24_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s24_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s24_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s24_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s24_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s24) = G_LOAD %0 :: (load (s24), align 1, addrspace 1)
%2:_(s32) = G_ANYEXT %1
@@ -1057,6 +1927,60 @@ body: |
; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
; GFX9-MESA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s48_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s48_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s48_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX11-HSA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s48_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX11-MESA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s48_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX12-HSA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s48_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
+ ; GFX12-MESA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s48) = G_LOAD %0 :: (load (s48), align 8, addrspace 1)
%2:_(s64) = G_ZEXT %1
@@ -1110,6 +2034,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s64_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s64_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s64_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s64_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s64_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s64_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 8, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -1162,6 +2128,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s64_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s64_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s64_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s64_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s64_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s64_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -1286,6 +2294,102 @@ body: |
; GFX9-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s64_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s64_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s64_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s64_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s64_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s64_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 2, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -1482,6 +2586,156 @@ body: |
; GFX9-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
; GFX9-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s64_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s64_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s64_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s64_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s64_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s64_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s64), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -1542,6 +2796,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s96_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s96_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s96_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s96_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s96_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s96_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -1605,6 +2907,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s96_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s96_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s96_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s96_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s96_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s96_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -1668,6 +3018,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s96_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s96_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s96_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s96_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s96_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s96_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -1810,6 +3208,117 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s96_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s96_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s96_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s96_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s96_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s96_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 2, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -2056,6 +3565,195 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s96_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s96_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s96_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s96_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s96_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s96_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = G_LOAD %0 :: (load (s96), align 1, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -2144,6 +3842,84 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; GFX9-MESA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s160_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
+ ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX10-HSA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s160_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX10-MESA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s160_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
+ ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX11-HSA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s160_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX11-MESA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s160_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
+ ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX12-HSA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s160_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
+ ; GFX12-MESA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s160) = G_LOAD %0 :: (load (s160), align 4, addrspace 1)
S_NOP 0, implicit %1
@@ -2253,6 +4029,102 @@ body: |
; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
; GFX9-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s224_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s224_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s224_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s224_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s224_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s224_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 1)
%2:_(s256) = G_IMPLICIT_DEF
@@ -2314,6 +4186,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s128_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s128_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s128_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s128_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s128_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s128_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -2372,6 +4292,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s128_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s128_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s128_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s128_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s128_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s128_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -2678,6 +4646,240 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s128_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s128_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s128_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s128_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX11-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s128_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s128_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s128), align 1, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -2736,6 +4938,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s256_align32
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s256_align32
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s256_align32
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s256_align32
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s256_align32
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s256_align32
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = G_LOAD %0 :: (load (s256), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -2788,6 +5038,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p1_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p1_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p1_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p1_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p1_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p1_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p1) = G_LOAD %0 :: (load (p1), align 8, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -2840,6 +5132,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p1_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p1_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p1_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p1_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p1_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p1_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p1) = G_LOAD %0 :: (load (p1), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -3040,6 +5374,159 @@ body: |
; GFX9-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p1_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p1_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p1_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p1_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p1_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p1_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p1) = G_LOAD %0 :: (load (p1), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -3092,6 +5579,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p3_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p3_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p3_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p3_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p3_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p3_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p3) = G_LOAD %0 :: (load (p3), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -3144,6 +5673,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p4_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p4_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p4_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p4_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p4_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p4_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 8, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -3196,6 +5767,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p4_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p4_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p4_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p4_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p4_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p4_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -3324,6 +5937,105 @@ body: |
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p4_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p4_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p4_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p4_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p4_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p4_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 2, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -3524,6 +6236,159 @@ body: |
; GFX9-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p4_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p4_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p4_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p4_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p4_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p4_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p4) = G_LOAD %0 :: (load (p4), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -3576,6 +6441,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p5_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p5_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p5_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p5_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p5_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p5_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p5) = G_LOAD %0 :: (load (p5), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -3656,6 +6563,69 @@ body: |
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p5_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p5_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p5_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p5_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p5_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p5_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p5) = G_LOAD %0 :: (load (p5), align 2, addrspace 1)
$vgpr0 = COPY %1
@@ -3776,6 +6746,99 @@ body: |
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_p5_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_p5_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_p5_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_p5_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_p5_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_p5_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p5) = G_LOAD %0 :: (load (p5), align 1, addrspace 1)
$vgpr0 = COPY %1
@@ -3828,6 +6891,80 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s8_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s8_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 4, addrspace 1)
%2:_(s16) = G_BITCAST %1
@@ -3882,6 +7019,80 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s8_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s8_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 2, addrspace 1)
%2:_(s16) = G_BITCAST %1
@@ -3960,6 +7171,82 @@ body: |
; GFX9-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s8_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s8_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s8_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s8_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 1, addrspace 1)
%2:_(s16) = G_BITCAST %1
@@ -4158,6 +7445,186 @@ body: |
; GFX9-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s8_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX10-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX10-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX10-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX10-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX10-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s8_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s8_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s8_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s8_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s8_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 4, addrspace 1)
%2:_(s24) = G_BITCAST %1
@@ -4407,6 +7874,231 @@ body: |
; GFX9-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
; GFX9-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s8_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX10-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX10-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX10-HSA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX10-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX10-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX10-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s8_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX10-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s8_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11-HSA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11-HSA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s8_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s8_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-HSA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-HSA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s8_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 1)
%2:_(s24) = G_BITCAST %1
@@ -4521,6 +8213,132 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s8_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s8_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s8_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s8_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s8_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s8_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -4653,6 +8471,147 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s8_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s8_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s8_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s8_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s8_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s8_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 2, addrspace 1)
$vgpr0 = COPY %1
@@ -4821,6 +8780,174 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s8_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s8_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C4]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s8_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s8_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C4]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s8_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s8_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C4]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 1, addrspace 1)
$vgpr0 = COPY %1
@@ -4873,6 +9000,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v8s8_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v8s8_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v8s8_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v8s8_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v8s8_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v8s8_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s8>) = G_LOAD %0 :: (load (<8 x s8>), align 8, addrspace 1)
%2:_(<2 x s32>) = G_BITCAST %1
@@ -4926,6 +9095,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v16s8_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v16s8_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v16s8_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v16s8_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v16s8_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v16s8_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<16 x s8>) = G_LOAD %0 :: (load (<16 x s8>), align 16, addrspace 1)
%2:_(<4 x s32>) = G_BITCAST %1
@@ -4979,6 +9190,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v32s8_align32
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v32s8_align32
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v32s8_align32
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v32s8_align32
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v32s8_align32
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v32s8_align32
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<32 x s8>) = G_LOAD %0 :: (load (<32 x s8>), align 32, addrspace 1)
%2:_(<8 x s32>) = G_BITCAST %1
@@ -5033,6 +9286,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s16_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s16_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s16_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s16_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -5121,6 +9416,88 @@ body: |
; GFX9-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s16_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v2s16_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v2s16_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s16_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v2s16_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v2s16_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 1)
$vgpr0 = COPY %1
@@ -5249,6 +9626,96 @@ body: |
; GFX9-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s16_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s16_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 1, addrspace 1)
$vgpr0 = COPY %1
@@ -5337,6 +9804,72 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s16_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s16_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s16_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s16_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 1)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -5507,6 +10040,204 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 4, addrspace 1)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -5677,6 +10408,204 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 2, addrspace 1)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -5903,6 +10832,210 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align1
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align1
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align1
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align1
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 1, addrspace 1)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -5957,6 +11090,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s16_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s16_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s16_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s16_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 8, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -6009,6 +11184,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s16_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s16_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s16_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s16_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -6143,6 +11360,134 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s16_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v4s16_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v4s16_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s16_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v4s16_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v4s16_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 2, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -6349,6 +11694,150 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s16_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s16_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -6429,6 +11918,84 @@ body: |
; GFX9-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v5s16_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v5s16_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v5s16_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v5s16_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s16>) = G_LOAD %0 :: (load (<5 x s16>), align 16, addrspace 1)
%2:_(<5 x s16>) = G_IMPLICIT_DEF
@@ -6639,6 +12206,286 @@ body: |
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s16>) = G_LOAD %0 :: (load (<5 x s16>), align 8, addrspace 1)
%2:_(<5 x s16>) = G_IMPLICIT_DEF
@@ -6849,6 +12696,286 @@ body: |
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s16>) = G_LOAD %0 :: (load (<5 x s16>), align 4, addrspace 1)
%2:_(<5 x s16>) = G_IMPLICIT_DEF
@@ -7077,6 +13204,286 @@ body: |
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s16>) = G_LOAD %0 :: (load (<5 x s16>), align 2, addrspace 1)
%2:_(<5 x s16>) = G_IMPLICIT_DEF
@@ -7393,6 +13800,302 @@ body: |
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align1
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align1
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v5s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX11-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align1
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align1
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v5s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX12-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s16>) = G_LOAD %0 :: (load (<5 x s16>), align 1, addrspace 1)
%2:_(<5 x s16>) = G_IMPLICIT_DEF
@@ -7459,6 +14162,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v6s16_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v6s16_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v6s16_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v6s16_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -7522,6 +14273,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v6s16_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v6s16_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v6s16_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v6s16_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -7585,6 +14384,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v6s16_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v6s16_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v6s16_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v6s16_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -7727,6 +14574,117 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v6s16_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v6s16_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v6s16_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v6s16_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 2, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -7973,6 +14931,195 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v6s16_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v6s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v6s16_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v6s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 1, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -8061,6 +15208,96 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-HSA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-MESA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v7s16_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v7s16_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v7s16_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v7s16_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-MESA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<7 x s16>) = G_LOAD %0 :: (load (<7 x s16>), align 16, addrspace 1)
%2:_(<7 x s16>) = G_IMPLICIT_DEF
@@ -8358,6 +15595,378 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<7 x s16>) = G_LOAD %0 :: (load (<7 x s16>), align 8, addrspace 1)
%2:_(<7 x s16>) = G_IMPLICIT_DEF
@@ -8655,6 +16264,378 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<7 x s16>) = G_LOAD %0 :: (load (<7 x s16>), align 4, addrspace 1)
%2:_(<7 x s16>) = G_IMPLICIT_DEF
@@ -8952,6 +16933,378 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<7 x s16>) = G_LOAD %0 :: (load (<7 x s16>), align 2, addrspace 1)
%2:_(<7 x s16>) = G_IMPLICIT_DEF
@@ -9369,6 +17722,402 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[OR6]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align1
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align1
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v7s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX11-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[OR6]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align1
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align1
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v7s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX12-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[OR6]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<7 x s16>) = G_LOAD %0 :: (load (<7 x s16>), align 1, addrspace 1)
%2:_(<7 x s16>) = G_IMPLICIT_DEF
@@ -9434,6 +18183,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v8s16_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v8s16_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v8s16_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v8s16_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v8s16_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v8s16_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -9492,6 +18289,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v8s16_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v8s16_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v8s16_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v8s16_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v8s16_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v8s16_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -9544,6 +18389,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s32_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s32_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s32_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s32_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 8, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -9596,6 +18483,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s32_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s32_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s32_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s32_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -9704,6 +18633,90 @@ body: |
; GFX9-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s32_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s32_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s32_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s32_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 2, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -9884,6 +18897,144 @@ body: |
; GFX9-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s32_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s32_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s32_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s32_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -9938,6 +19089,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s32_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s32_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s32_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s32_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s32_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s32_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -9995,6 +19188,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s32_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s32_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s32_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s32_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s32_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s32_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -10047,6 +19282,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s32_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s32_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s32_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s32_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s32_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s32_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -10099,6 +19376,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s32_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s32_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s32_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s32_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s32_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s32_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -10151,6 +19470,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s32_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s32_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s32_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s32_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s32_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s32_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -10203,6 +19564,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v8s32_align32
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v8s32_align32
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v8s32_align32
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v8s32_align32
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v8s32_align32
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v8s32_align32
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = G_LOAD %0 :: (load (<8 x s32>), align 32, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -10255,6 +19658,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v16s32_align32
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v16s32_align32
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v16s32_align32
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v16s32_align32
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v16s32_align32
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v16s32_align32
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<16 x s32>) = G_LOAD %0 :: (load (<16 x s32>), align 32, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
@@ -10307,6 +19752,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s64_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s64_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s64_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s64_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -10359,6 +19846,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s64_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s64_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s64_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s64_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -10411,6 +19940,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s64_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s64_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s64_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s64_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -10611,6 +20182,159 @@ body: |
; GFX9-MESA-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s64_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s64_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s64_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s64_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 2, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -10947,6 +20671,261 @@ body: |
; GFX9-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s64_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s64_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX11-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX11-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX11-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s64_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s64_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX12-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX12-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX12-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX12-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 1, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -11005,6 +20984,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2sp1_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2sp1_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2sp1_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2sp1_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2sp1_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2sp1_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -11057,6 +21084,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s64_align32
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s64_align32
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s64_align32
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s64_align32
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s64_align32
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s64_align32
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 1)
%2:_(<4 x s64>) = G_IMPLICIT_DEF
@@ -11147,6 +21216,84 @@ body: |
; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s64_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
+ ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s64_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s64_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
+ ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s64_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s64_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
+ ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s64_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 8, addrspace 1)
%2:_(<4 x s64>) = G_IMPLICIT_DEF
@@ -11637,6 +21784,384 @@ body: |
; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v3s64_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v3s64_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX10-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX10-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX10-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX10-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX10-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX10-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v3s64_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v3s64_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX11-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX11-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX11-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX11-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX11-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX11-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX11-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX11-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX11-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX11-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX11-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v3s64_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v3s64_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX12-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX12-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX12-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX12-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX12-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX12-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX12-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX12-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX12-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX12-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX12-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX12-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 1, addrspace 1)
%2:_(<4 x s64>) = G_IMPLICIT_DEF
@@ -11691,6 +22216,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s64_align32
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s64_align32
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s64_align32
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s64_align32
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s64_align32
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s64_align32
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 32, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -11743,6 +22310,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s64_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s64_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s64_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s64_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s64_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s64_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -12351,6 +22960,465 @@ body: |
; GFX9-MESA-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4s64_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4s64_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX10-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX10-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX10-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX10-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX10-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX10-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX10-MESA-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
+ ; GFX10-MESA-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C8]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p1) :: (load (s8) from unknown-address + 24, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p1) :: (load (s8) from unknown-address + 25, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p1) :: (load (s8) from unknown-address + 26, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p1) :: (load (s8) from unknown-address + 27, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
+ ; GFX10-MESA-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
+ ; GFX10-MESA-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
+ ; GFX10-MESA-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p1) :: (load (s8) from unknown-address + 28, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p1) :: (load (s8) from unknown-address + 29, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p1) :: (load (s8) from unknown-address + 30, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p1) :: (load (s8) from unknown-address + 31, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
+ ; GFX10-MESA-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
+ ; GFX10-MESA-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
+ ; GFX10-MESA-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
+ ; GFX10-MESA-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4s64_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4s64_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX11-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX11-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX11-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX11-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX11-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX11-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX11-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX11-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX11-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX11-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX11-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX11-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX11-MESA-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
+ ; GFX11-MESA-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C8]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p1) :: (load (s8) from unknown-address + 24, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p1) :: (load (s8) from unknown-address + 25, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p1) :: (load (s8) from unknown-address + 26, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p1) :: (load (s8) from unknown-address + 27, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
+ ; GFX11-MESA-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
+ ; GFX11-MESA-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
+ ; GFX11-MESA-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p1) :: (load (s8) from unknown-address + 28, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p1) :: (load (s8) from unknown-address + 29, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p1) :: (load (s8) from unknown-address + 30, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p1) :: (load (s8) from unknown-address + 31, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
+ ; GFX11-MESA-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
+ ; GFX11-MESA-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
+ ; GFX11-MESA-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
+ ; GFX11-MESA-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4s64_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4s64_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX12-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX12-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX12-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX12-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX12-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
+ ; GFX12-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
+ ; GFX12-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
+ ; GFX12-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
+ ; GFX12-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
+ ; GFX12-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
+ ; GFX12-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
+ ; GFX12-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
+ ; GFX12-MESA-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
+ ; GFX12-MESA-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C8]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p1) :: (load (s8) from unknown-address + 24, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p1) :: (load (s8) from unknown-address + 25, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p1) :: (load (s8) from unknown-address + 26, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p1) :: (load (s8) from unknown-address + 27, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
+ ; GFX12-MESA-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
+ ; GFX12-MESA-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
+ ; GFX12-MESA-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p1) :: (load (s8) from unknown-address + 28, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p1) :: (load (s8) from unknown-address + 29, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p1) :: (load (s8) from unknown-address + 30, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p1) :: (load (s8) from unknown-address + 31, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
+ ; GFX12-MESA-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
+ ; GFX12-MESA-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
+ ; GFX12-MESA-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
+ ; GFX12-MESA-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 1, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -12409,6 +23477,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2s128_align32
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2s128_align32
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2s128_align32
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2s128_align32
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2s128_align32
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2s128_align32
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s128>) = G_LOAD %0 :: (load (<2 x s128>), align 32, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -12467,6 +23583,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2p1_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2p1_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2p1_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2p1_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 16, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12525,6 +23689,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2p1_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2p1_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2p1_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2p1_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12583,6 +23795,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2p1_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2p1_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2p1_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2p1_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12889,6 +24149,240 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2p1_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2p1_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX11-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2p1_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2p1_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 1, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12947,6 +24441,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v4p1_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v4p1_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v4p1_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v4p1_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v4p1_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v4p1_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x p1>) = G_LOAD %0 :: (load (<4 x p1>), align 8, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -13005,6 +24547,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2p3_align8
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2p3_align8
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2p3_align8
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2p3_align8
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2p3_align8
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2p3_align8
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 8, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -13063,6 +24653,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2p3_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2p3_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2p3_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2p3_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2p3_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2p3_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -13249,6 +24887,150 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v2p3_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v2p3_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v2p3_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v2p3_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v2p3_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v2p3_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -13301,6 +25083,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_1_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_1_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_1_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s32_from_1_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s32_from_1_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s32_from_1_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s8), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -13353,6 +25177,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_2_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_2_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_2_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s32_from_2_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s32_from_2_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s32_from_2_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s16), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -13465,6 +25331,102 @@ body: |
; GFX9-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
; GFX9-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s24), align 1, addrspace 1)
$vgpr0 = COPY %1
@@ -13552,6 +25514,84 @@ body: |
; GFX9-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s24), align 2, addrspace 1)
$vgpr0 = COPY %1
@@ -13604,6 +25644,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s24), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -13663,6 +25745,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_1_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_1_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s64_from_1_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s64_from_1_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s64_from_1_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s64_from_1_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -13721,6 +25851,86 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -13779,6 +25989,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_4_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_4_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s64_from_4_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s64_from_4_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s64_from_4_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s64_from_4_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s32), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -13855,6 +26113,72 @@ body: |
; GFX9-MESA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX9-MESA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s128_from_4_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX10-HSA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-HSA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s128_from_4_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX10-MESA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX10-MESA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s128_from_4_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX11-HSA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s128_from_4_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX11-MESA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s128_from_4_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX12-HSA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s128_from_4_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX12-MESA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = G_LOAD %0 :: (load (s32), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -13913,6 +26237,86 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-HSA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -13971,6 +26375,54 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_1_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_1_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-HSA-LABEL: name: test_ext_load_global_s64_from_1_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-MESA-LABEL: name: test_ext_load_global_s64_from_1_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-HSA-LABEL: name: test_ext_load_global_s64_from_1_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-MESA-LABEL: name: test_ext_load_global_s64_from_1_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -14023,6 +26475,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s16>), align 1, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -14075,6 +26569,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -14127,6 +26663,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s16>), align 4, addrspace 1)
$vgpr0_vgpr1 = COPY %1
@@ -14179,6 +26757,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s16>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -14231,6 +26851,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s16>), align 4, addrspace 1)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -14683,6 +27345,360 @@ body: |
; GFX9-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_global_v2s96_align1
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_global_v2s96_align1
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
+ ; GFX10-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
+ ; GFX10-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_global_v2s96_align1
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_global_v2s96_align1
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX11-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX11-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
+ ; GFX11-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
+ ; GFX11-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_global_v2s96_align1
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_global_v2s96_align1
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
+ ; GFX12-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
+ ; GFX12-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 1, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -14938,6 +27954,210 @@ body: |
; GFX9-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_global_v2s96_align2
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_global_v2s96_align2
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s16) from unknown-address + 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s16) from unknown-address + 18, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s64)
+ ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s16) from unknown-address + 20, addrspace 1)
+ ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s16) from unknown-address + 22, addrspace 1)
+ ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
+ ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_global_v2s96_align2
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_global_v2s96_align2
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX11-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s16) from unknown-address + 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s16) from unknown-address + 18, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX11-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s64)
+ ; GFX11-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s16) from unknown-address + 20, addrspace 1)
+ ; GFX11-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s16) from unknown-address + 22, addrspace 1)
+ ; GFX11-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
+ ; GFX11-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_global_v2s96_align2
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_global_v2s96_align2
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s16) from unknown-address + 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s16) from unknown-address + 18, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX12-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s64)
+ ; GFX12-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s16) from unknown-address + 20, addrspace 1)
+ ; GFX12-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s16) from unknown-address + 22, addrspace 1)
+ ; GFX12-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
+ ; GFX12-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 2, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -15050,6 +28270,96 @@ body: |
; GFX9-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_global_v2s96_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_global_v2s96_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_global_v2s96_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_global_v2s96_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_global_v2s96_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_global_v2s96_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 4, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -15160,6 +28470,96 @@ body: |
; GFX9-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-HSA-LABEL: name: test_global_v2s96_align16
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX10-MESA-LABEL: name: test_global_v2s96_align16
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-HSA-LABEL: name: test_global_v2s96_align16
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX11-MESA-LABEL: name: test_global_v2s96_align16
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX11-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX11-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX11-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX11-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-HSA-LABEL: name: test_global_v2s96_align16
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-MESA-LABEL: name: test_global_v2s96_align16
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 16, addrspace 1)
%2:_(s96) = G_EXTRACT %1, 0
@@ -15695,6 +29095,720 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
; GFX9-MESA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v32s1_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX10-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
+ ; GFX10-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX10-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX10-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
+ ; GFX10-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX10-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-HSA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
+ ; GFX10-HSA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
+ ; GFX10-HSA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
+ ; GFX10-HSA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+ ; GFX10-HSA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
+ ; GFX10-HSA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
+ ; GFX10-HSA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
+ ; GFX10-HSA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX10-HSA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
+ ; GFX10-HSA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
+ ; GFX10-HSA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
+ ; GFX10-HSA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
+ ; GFX10-HSA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
+ ; GFX10-HSA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
+ ; GFX10-HSA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
+ ; GFX10-HSA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
+ ; GFX10-HSA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
+ ; GFX10-HSA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
+ ; GFX10-HSA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
+ ; GFX10-HSA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
+ ; GFX10-HSA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
+ ; GFX10-HSA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
+ ; GFX10-HSA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX10-HSA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
+ ; GFX10-HSA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
+ ; GFX10-HSA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
+ ; GFX10-HSA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
+ ; GFX10-HSA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
+ ; GFX10-HSA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
+ ; GFX10-HSA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
+ ; GFX10-HSA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-HSA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
+ ; GFX10-HSA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
+ ; GFX10-HSA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
+ ; GFX10-HSA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
+ ; GFX10-HSA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
+ ; GFX10-HSA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
+ ; GFX10-HSA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
+ ; GFX10-HSA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX10-HSA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
+ ; GFX10-HSA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
+ ; GFX10-HSA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
+ ; GFX10-HSA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
+ ; GFX10-HSA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
+ ; GFX10-HSA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
+ ; GFX10-HSA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX10-HSA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v32s1_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX10-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
+ ; GFX10-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX10-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
+ ; GFX10-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
+ ; GFX10-MESA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
+ ; GFX10-MESA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
+ ; GFX10-MESA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+ ; GFX10-MESA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
+ ; GFX10-MESA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
+ ; GFX10-MESA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
+ ; GFX10-MESA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX10-MESA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
+ ; GFX10-MESA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
+ ; GFX10-MESA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
+ ; GFX10-MESA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
+ ; GFX10-MESA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
+ ; GFX10-MESA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
+ ; GFX10-MESA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
+ ; GFX10-MESA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
+ ; GFX10-MESA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
+ ; GFX10-MESA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
+ ; GFX10-MESA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
+ ; GFX10-MESA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
+ ; GFX10-MESA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
+ ; GFX10-MESA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
+ ; GFX10-MESA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX10-MESA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
+ ; GFX10-MESA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
+ ; GFX10-MESA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
+ ; GFX10-MESA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
+ ; GFX10-MESA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
+ ; GFX10-MESA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
+ ; GFX10-MESA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
+ ; GFX10-MESA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-MESA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
+ ; GFX10-MESA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
+ ; GFX10-MESA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
+ ; GFX10-MESA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
+ ; GFX10-MESA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
+ ; GFX10-MESA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
+ ; GFX10-MESA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
+ ; GFX10-MESA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX10-MESA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
+ ; GFX10-MESA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
+ ; GFX10-MESA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
+ ; GFX10-MESA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
+ ; GFX10-MESA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
+ ; GFX10-MESA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
+ ; GFX10-MESA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX10-MESA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v32s1_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX11-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX11-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
+ ; GFX11-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX11-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX11-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX11-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
+ ; GFX11-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX11-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
+ ; GFX11-HSA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
+ ; GFX11-HSA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
+ ; GFX11-HSA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+ ; GFX11-HSA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
+ ; GFX11-HSA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
+ ; GFX11-HSA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
+ ; GFX11-HSA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX11-HSA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
+ ; GFX11-HSA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
+ ; GFX11-HSA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
+ ; GFX11-HSA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
+ ; GFX11-HSA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
+ ; GFX11-HSA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
+ ; GFX11-HSA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
+ ; GFX11-HSA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
+ ; GFX11-HSA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
+ ; GFX11-HSA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
+ ; GFX11-HSA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
+ ; GFX11-HSA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
+ ; GFX11-HSA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
+ ; GFX11-HSA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
+ ; GFX11-HSA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX11-HSA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
+ ; GFX11-HSA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
+ ; GFX11-HSA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
+ ; GFX11-HSA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
+ ; GFX11-HSA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
+ ; GFX11-HSA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
+ ; GFX11-HSA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
+ ; GFX11-HSA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-HSA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
+ ; GFX11-HSA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
+ ; GFX11-HSA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
+ ; GFX11-HSA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
+ ; GFX11-HSA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
+ ; GFX11-HSA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
+ ; GFX11-HSA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
+ ; GFX11-HSA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX11-HSA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
+ ; GFX11-HSA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
+ ; GFX11-HSA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
+ ; GFX11-HSA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
+ ; GFX11-HSA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
+ ; GFX11-HSA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
+ ; GFX11-HSA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX11-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX11-HSA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v32s1_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX11-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
+ ; GFX11-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX11-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
+ ; GFX11-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX11-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
+ ; GFX11-MESA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
+ ; GFX11-MESA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
+ ; GFX11-MESA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+ ; GFX11-MESA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
+ ; GFX11-MESA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
+ ; GFX11-MESA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
+ ; GFX11-MESA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX11-MESA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
+ ; GFX11-MESA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
+ ; GFX11-MESA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
+ ; GFX11-MESA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
+ ; GFX11-MESA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
+ ; GFX11-MESA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
+ ; GFX11-MESA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
+ ; GFX11-MESA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
+ ; GFX11-MESA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
+ ; GFX11-MESA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
+ ; GFX11-MESA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
+ ; GFX11-MESA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
+ ; GFX11-MESA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
+ ; GFX11-MESA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
+ ; GFX11-MESA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX11-MESA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
+ ; GFX11-MESA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
+ ; GFX11-MESA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
+ ; GFX11-MESA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
+ ; GFX11-MESA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
+ ; GFX11-MESA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
+ ; GFX11-MESA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
+ ; GFX11-MESA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-MESA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
+ ; GFX11-MESA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
+ ; GFX11-MESA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
+ ; GFX11-MESA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
+ ; GFX11-MESA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
+ ; GFX11-MESA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
+ ; GFX11-MESA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
+ ; GFX11-MESA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX11-MESA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
+ ; GFX11-MESA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
+ ; GFX11-MESA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
+ ; GFX11-MESA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
+ ; GFX11-MESA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
+ ; GFX11-MESA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
+ ; GFX11-MESA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX11-MESA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v32s1_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX12-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX12-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
+ ; GFX12-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX12-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX12-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX12-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
+ ; GFX12-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX12-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
+ ; GFX12-HSA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
+ ; GFX12-HSA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
+ ; GFX12-HSA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+ ; GFX12-HSA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
+ ; GFX12-HSA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
+ ; GFX12-HSA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
+ ; GFX12-HSA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-HSA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
+ ; GFX12-HSA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
+ ; GFX12-HSA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
+ ; GFX12-HSA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
+ ; GFX12-HSA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
+ ; GFX12-HSA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
+ ; GFX12-HSA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
+ ; GFX12-HSA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
+ ; GFX12-HSA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
+ ; GFX12-HSA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
+ ; GFX12-HSA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
+ ; GFX12-HSA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
+ ; GFX12-HSA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
+ ; GFX12-HSA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
+ ; GFX12-HSA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX12-HSA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
+ ; GFX12-HSA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
+ ; GFX12-HSA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
+ ; GFX12-HSA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
+ ; GFX12-HSA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
+ ; GFX12-HSA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
+ ; GFX12-HSA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
+ ; GFX12-HSA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-HSA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
+ ; GFX12-HSA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
+ ; GFX12-HSA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
+ ; GFX12-HSA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
+ ; GFX12-HSA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
+ ; GFX12-HSA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
+ ; GFX12-HSA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
+ ; GFX12-HSA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX12-HSA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
+ ; GFX12-HSA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
+ ; GFX12-HSA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
+ ; GFX12-HSA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
+ ; GFX12-HSA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
+ ; GFX12-HSA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
+ ; GFX12-HSA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX12-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX12-HSA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v32s1_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX12-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
+ ; GFX12-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX12-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
+ ; GFX12-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX12-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
+ ; GFX12-MESA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
+ ; GFX12-MESA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
+ ; GFX12-MESA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+ ; GFX12-MESA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
+ ; GFX12-MESA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
+ ; GFX12-MESA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
+ ; GFX12-MESA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-MESA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
+ ; GFX12-MESA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
+ ; GFX12-MESA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
+ ; GFX12-MESA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
+ ; GFX12-MESA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
+ ; GFX12-MESA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
+ ; GFX12-MESA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
+ ; GFX12-MESA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
+ ; GFX12-MESA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
+ ; GFX12-MESA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
+ ; GFX12-MESA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
+ ; GFX12-MESA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
+ ; GFX12-MESA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
+ ; GFX12-MESA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
+ ; GFX12-MESA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX12-MESA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
+ ; GFX12-MESA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
+ ; GFX12-MESA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
+ ; GFX12-MESA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
+ ; GFX12-MESA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
+ ; GFX12-MESA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
+ ; GFX12-MESA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
+ ; GFX12-MESA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-MESA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
+ ; GFX12-MESA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
+ ; GFX12-MESA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
+ ; GFX12-MESA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
+ ; GFX12-MESA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
+ ; GFX12-MESA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
+ ; GFX12-MESA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
+ ; GFX12-MESA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX12-MESA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
+ ; GFX12-MESA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
+ ; GFX12-MESA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
+ ; GFX12-MESA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
+ ; GFX12-MESA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
+ ; GFX12-MESA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
+ ; GFX12-MESA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
+ ; GFX12-MESA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<32 x s1>) = G_LOAD %0 :: (load (<32 x s1>), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -15867,6 +29981,216 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
; GFX9-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_v8s4_align4
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX10-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX10-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX10-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX10-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX10-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_v8s4_align4
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX10-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX10-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX10-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX10-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX10-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_v8s4_align4
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX11-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX11-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX11-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX11-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX11-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX11-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX11-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX11-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX11-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_v8s4_align4
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX11-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX11-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX11-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX11-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX11-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX11-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX11-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX11-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_v8s4_align4
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX12-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX12-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX12-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX12-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX12-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX12-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX12-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX12-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_v8s4_align4
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
+ ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
+ ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+ ; GFX12-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
+ ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+ ; GFX12-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
+ ; GFX12-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
+ ; GFX12-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
+ ; GFX12-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
+ ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX12-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s4>) = G_LOAD %0 :: (load (<8 x s4>), align 4, addrspace 1)
$vgpr0 = COPY %1
@@ -15920,6 +30244,48 @@ body: |
; GFX9-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-HSA-LABEL: name: test_load_global_s32_align536870912
+ ; GFX10-HSA: liveins: $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: {{ $}}
+ ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
+ ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX10-MESA-LABEL: name: test_load_global_s32_align536870912
+ ; GFX10-MESA: liveins: $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: {{ $}}
+ ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
+ ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-HSA-LABEL: name: test_load_global_s32_align536870912
+ ; GFX11-HSA: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: {{ $}}
+ ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
+ ; GFX11-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-MESA-LABEL: name: test_load_global_s32_align536870912
+ ; GFX11-MESA: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: {{ $}}
+ ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
+ ; GFX11-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-HSA-LABEL: name: test_load_global_s32_align536870912
+ ; GFX12-HSA: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: {{ $}}
+ ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
+ ; GFX12-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-MESA-LABEL: name: test_load_global_s32_align536870912
+ ; GFX12-MESA: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: {{ $}}
+ ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
+ ; GFX12-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = G_LOAD %0 :: (load (s16), align 536870912, addrspace 1)
$vgpr0 = COPY %1
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir
index a38d0549d22e1..f369d588b509b 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir
@@ -7,10 +7,14 @@
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9-UNALIGNED %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX10 %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX10-UNALIGNED %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11 %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-UNALIGNED %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11 %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-UNALIGNED %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-UNALIGNED,GFX11-UNALIGNED-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-UNALIGNED,GFX11-UNALIGNED-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12,GFX12-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12-UNALIGNED,GFX12-UNALIGNED-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12-UNALIGNED,GFX12-UNALIGNED-FAKE16 %s
---
name: test_load_local_s1_align1
@@ -107,6 +111,24 @@ body: |
; GFX11-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX11-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s1_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s1_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[AND]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s1) = G_LOAD %0 :: (load (s1), align 1, addrspace 3)
%2:_(s32) = G_ZEXT %1
@@ -208,6 +230,24 @@ body: |
; GFX11-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
; GFX11-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s2_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[AND]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s2_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
+ ; GFX12-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[AND]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s2) = G_LOAD %0 :: (load (s2), align 1, addrspace 3)
%2:_(s32) = G_ZEXT %1
@@ -289,6 +329,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s8_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s8_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s8) = G_LOAD %0 :: (load (s8), align 4, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -370,6 +424,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s8_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s8_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s8) = G_LOAD %0 :: (load (s8), align 1, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -438,19 +506,65 @@ body: |
; GFX10-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
; GFX10-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-LABEL: name: test_load_local_s16_align4
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
- ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_s16_align4
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_s16_align4
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_s16_align4
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_s16_align4
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_s16_align4
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_s16_align4
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_s16_align4
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_s16_align4
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_s16_align4
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s16) = G_LOAD %0 :: (load (s16), align 4, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -519,19 +633,65 @@ body: |
; GFX10-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX10-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-LABEL: name: test_load_local_s16_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_s16_align2
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_s16_align2
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_s16_align2
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_s16_align2
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_s16_align2
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s16) = G_LOAD %0 :: (load (s16), align 2, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -649,12 +809,48 @@ body: |
; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX11-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_s16_align1
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_s16_align1
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_s16_align1
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s16_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_s16_align1
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_s16_align1
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s16) = G_LOAD %0 :: (load (s16), align 1, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -736,6 +932,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s32_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s32_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s32) = G_LOAD %0 :: (load (s32), align 4, addrspace 3)
$vgpr0 = COPY %1
@@ -858,6 +1068,26 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s32_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s32_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s32) = G_LOAD %0 :: (load (s32), align 2, addrspace 3)
$vgpr0 = COPY %1
@@ -1050,6 +1280,36 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s32_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s32_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s32) = G_LOAD %0 :: (load (s32), align 1, addrspace 3)
$vgpr0 = COPY %1
@@ -1130,6 +1390,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 8, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s24_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 8, addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s24_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 8, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s24) = G_LOAD %0 :: (load (s24), align 8, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -1211,6 +1485,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s24_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s24_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s24) = G_LOAD %0 :: (load (s24), align 4, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -1352,6 +1640,32 @@ body: |
; GFX11-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX11-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s24_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, align 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s24_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s24) = G_LOAD %0 :: (load (s24), align 2, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -1535,6 +1849,38 @@ body: |
; GFX11-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
; GFX11-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s24_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[OR1]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s24_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s24) = G_LOAD %0 :: (load (s24), align 1, addrspace 3)
%2:_(s32) = G_ANYEXT %1
@@ -1616,6 +1962,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s48_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s48_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s48) = G_LOAD %0 :: (load (s48), align 8, addrspace 3)
%2:_(s64) = G_ANYEXT %1
@@ -1697,6 +2057,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s64_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s64_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s64), align 8, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -1777,6 +2151,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s64_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s64_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s64), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -1991,6 +2379,38 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s64_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s64_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s64), align 2, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -2331,6 +2751,56 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s64_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s64_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s64), align 1, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -2750,6 +3220,68 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s96_align16
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s96_align16
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(s96) = G_LOAD %0 :: (load (s96), align 1, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -2892,6 +3424,29 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 8, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s96_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), align 8, addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 8, align 8, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s96_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 8, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(s96) = G_LOAD %0 :: (load (s96), align 8, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -3034,6 +3589,29 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s96_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s96_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -3278,6 +3856,43 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s96_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s16) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p3) :: (load (s16) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s96_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(s96) = G_LOAD %0 :: (load (s96), align 2, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -3697,6 +4312,68 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s96_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s96_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(s96) = G_LOAD %0 :: (load (s96), align 1, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -4226,6 +4903,83 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s128_align16
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p3) :: (load (s8) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p3) :: (load (s8) from unknown-address + 13, addrspace 3)
+ ; GFX12-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p3) :: (load (s8) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p3) :: (load (s8) from unknown-address + 15, addrspace 3)
+ ; GFX12-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s128_align16
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p3) = COPY $vgpr0
%1:_(s128) = G_LOAD %0 :: (load (s128), align 1, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4344,6 +5098,22 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s128_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s128_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p3) = COPY $vgpr0
%1:_(s128) = G_LOAD %0 :: (load (s128), align 8, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4502,6 +5272,32 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s128_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s128_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p3) = COPY $vgpr0
%1:_(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -4800,6 +5596,50 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s128_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s16) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p3) :: (load (s16) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s16) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s16) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32), [[OR3]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s128_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p3) = COPY $vgpr0
%1:_(s128) = G_LOAD %0 :: (load (s128), align 2, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -5329,6 +6169,83 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-LABEL: name: test_load_local_s128_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p3) :: (load (s8) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p3) :: (load (s8) from unknown-address + 13, addrspace 3)
+ ; GFX12-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p3) :: (load (s8) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p3) :: (load (s8) from unknown-address + 15, addrspace 3)
+ ; GFX12-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_s128_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
%0:_(p3) = COPY $vgpr0
%1:_(s128) = G_LOAD %0 :: (load (s128), align 1, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -5409,6 +6326,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p1_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p1_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p3) = COPY $vgpr0
%1:_(p1) = G_LOAD %0 :: (load (p1), align 8, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -5516,6 +6447,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p1_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p1_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p3) = COPY $vgpr0
%1:_(p1) = G_LOAD %0 :: (load (p1), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -5738,6 +6683,39 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p1_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
+ ; GFX12-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR2]](s64)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p1_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p3) = COPY $vgpr0
%1:_(p1) = G_LOAD %0 :: (load (p1), align 2, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -6086,6 +7064,57 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p1_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p1_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p3) :: (load (p1), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
%0:_(p3) = COPY $vgpr0
%1:_(p1) = G_LOAD %0 :: (load (p1), align 1, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -6166,6 +7195,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p3) :: (load (p3), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p3_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p3) :: (load (p3), addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p3_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p3) :: (load (p3), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p3)
%0:_(p3) = COPY $vgpr0
%1:_(p3) = G_LOAD %0 :: (load (p3), align 4, addrspace 3)
$vgpr0 = COPY %1
@@ -6295,6 +7338,27 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p3) :: (load (p3), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p3_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[INTTOPTR:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR]](s32)
+ ; GFX12-NEXT: $vgpr0 = COPY [[INTTOPTR]](p3)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p3_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p3) :: (load (p3), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p3)
%0:_(p3) = COPY $vgpr0
%1:_(p3) = G_LOAD %0 :: (load (p3), align 2, addrspace 3)
$vgpr0 = COPY %1
@@ -6494,6 +7558,37 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p3) :: (load (p3), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p3)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p3_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[INTTOPTR:%[0-9]+]]:_(p3) = G_INTTOPTR [[OR2]](s32)
+ ; GFX12-NEXT: $vgpr0 = COPY [[INTTOPTR]](p3)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p3_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p3) :: (load (p3), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p3)
%0:_(p3) = COPY $vgpr0
%1:_(p3) = G_LOAD %0 :: (load (p3), align 1, addrspace 3)
$vgpr0 = COPY %1
@@ -6574,6 +7669,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p3) :: (load (p5), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p5_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p3) :: (load (p5), addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p5_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p3) :: (load (p5), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p5)
%0:_(p3) = COPY $vgpr0
%1:_(p5) = G_LOAD %0 :: (load (p5), align 4, addrspace 3)
$vgpr0 = COPY %1
@@ -6703,6 +7812,27 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p3) :: (load (p5), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p5_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
+ ; GFX12-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p5_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p3) :: (load (p5), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p5)
%0:_(p3) = COPY $vgpr0
%1:_(p5) = G_LOAD %0 :: (load (p5), align 2, addrspace 3)
$vgpr0 = COPY %1
@@ -6902,6 +8032,37 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p3) :: (load (p5), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p5)
+ ;
+ ; GFX12-LABEL: name: test_load_local_p5_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
+ ; GFX12-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_p5_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p3) :: (load (p5), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](p5)
%0:_(p3) = COPY $vgpr0
%1:_(p5) = G_LOAD %0 :: (load (p5), align 1, addrspace 3)
$vgpr0 = COPY %1
@@ -6969,19 +8130,65 @@ body: |
; GFX10-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
; GFX10-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-LABEL: name: test_load_local_v2s8_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_v2s8_align2
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_v2s8_align2
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 2, addrspace 3)
%2:_(s16) = G_BITCAST %1
@@ -7120,15 +8327,64 @@ body: |
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[LSHR]](s32)
; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_v2s8_align1
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
- ; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_v2s8_align1
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[LOAD]], [[C]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_v2s8_align1
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s8_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[LSHR]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_v2s8_align1
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[LOAD]], [[C]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_v2s8_align1
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 1, addrspace 3)
%2:_(<2 x s32>) = G_ANYEXT %1
@@ -7446,6 +8702,66 @@ body: |
; GFX11-UNALIGNED-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
; GFX11-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v3s8_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s8_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-UNALIGNED-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-UNALIGNED-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-UNALIGNED-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), addrspace 1, align 4)
%2:_(s24) = G_BITCAST %1
@@ -7849,6 +9165,81 @@ body: |
; GFX11-UNALIGNED-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
; GFX11-UNALIGNED-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v3s8_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX12-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s8_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-UNALIGNED-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-UNALIGNED-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-UNALIGNED-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-UNALIGNED-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-UNALIGNED-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-UNALIGNED-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[OR3]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 3)
%2:_(s24) = G_BITCAST %1
@@ -7931,6 +9322,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s8_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s8_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s8>) = G_LOAD %0 :: (load (<4 x s8>), align 4, addrspace 3)
%2:_(s32) = G_BITCAST %1
@@ -8012,6 +9417,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v8s8_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v8s8_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<8 x s8>) = G_LOAD %0 :: (load (<8 x s8>), align 8, addrspace 3)
%2:_(<2 x s32>) = G_BITCAST %1
@@ -8528,6 +9947,81 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v16s8_align16
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p3) :: (load (s8) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p3) :: (load (s8) from unknown-address + 13, addrspace 3)
+ ; GFX12-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p3) :: (load (s8) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p3) :: (load (s8) from unknown-address + 15, addrspace 3)
+ ; GFX12-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v16s8_align16
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<16 x s8>) = G_LOAD %0 :: (load (<16 x s8>), align 1, addrspace 3)
%2:_(<4 x s32>) = G_BITCAST %1
@@ -8609,6 +10103,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s16_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s16_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 4, addrspace 3)
$vgpr0 = COPY %1
@@ -8728,18 +10236,29 @@ body: |
; GFX10-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 2, addrspace 3)
; GFX10-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
- ; GFX11-LABEL: name: test_load_local_v2s16_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_v2s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_v2s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
; GFX11-UNALIGNED-LABEL: name: test_load_local_v2s16_align2
; GFX11-UNALIGNED: liveins: $vgpr0
@@ -8747,6 +10266,37 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_v2s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_v2s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s16_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 3)
$vgpr0 = COPY %1
@@ -8955,6 +10505,36 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s16_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s16_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 1, addrspace 3)
$vgpr0 = COPY %1
@@ -9087,6 +10667,28 @@ body: |
; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
; GFX11-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v3s16_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s16_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 3)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -9300,47 +10902,161 @@ body: |
; GFX10-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-LABEL: name: test_load_local_v3s16_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_v3s16_align2
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX11-UNALIGNED-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s16_align2
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 2, addrspace 3)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -9673,26 +11389,118 @@ body: |
; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_v3s16_align1
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, align 1, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX11-UNALIGNED-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, align 1, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s16_align1
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, align 1, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, align 1, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s16_align1
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, align 1, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, align 1, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v3s16_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s16_align1
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, align 1, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, align 1, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s16_align1
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, align 1, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, align 1, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 1, addrspace 3)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -9774,6 +11582,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s16_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s16_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 8, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -9908,6 +11730,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s16_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s16_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -10110,28 +11946,47 @@ body: |
; GFX10-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX10-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; GFX11-LABEL: name: test_load_local_v4s16_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
- ; GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
- ; GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_v4s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_v4s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
; GFX11-UNALIGNED-LABEL: name: test_load_local_v4s16_align2
; GFX11-UNALIGNED: liveins: $vgpr0
@@ -10139,6 +11994,55 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_v4s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_v4s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s16_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 2, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -10497,6 +12401,54 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s16_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s16_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 1, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -10577,6 +12529,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s32_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s32_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 8, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -10657,6 +12623,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s32_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s32_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -10839,6 +12819,34 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s32_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s32_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 2, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -11147,6 +13155,52 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s32_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s32_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 1, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -11556,6 +13610,66 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v3s32_align16
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s32_align16
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 1, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -11688,6 +13802,27 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v3s32_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s32_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -11776,6 +13911,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s32_align16
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s32_align16
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 16, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -11884,6 +14033,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s32_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s32_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 8, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12032,6 +14195,30 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s32_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s32_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12320,6 +14507,48 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s32_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s16) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p3) :: (load (s16) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s16) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s16) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32), [[OR3]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s32_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 2, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12839,6 +15068,81 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s32_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p3) :: (load (s8) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p3) :: (load (s8) from unknown-address + 13, addrspace 3)
+ ; GFX12-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p3) :: (load (s8) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p3) :: (load (s8) from unknown-address + 15, addrspace 3)
+ ; GFX12-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s32_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 1, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -12971,6 +15275,28 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<4 x s32>) from unknown-address + 16, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v8s32_align32
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 32, addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<4 x s32>) from unknown-address + 16, addrspace 3)
+ ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v8s32_align32
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 32, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<4 x s32>) from unknown-address + 16, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<8 x s32>) = G_LOAD %0 :: (load (<8 x s32>), align 32, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -13175,6 +15501,40 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p3) :: (load (<4 x s32>) from unknown-address + 48, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v16s32_align32
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 32, addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<4 x s32>) from unknown-address + 16, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p3) :: (load (<4 x s32>) from unknown-address + 32, align 32, addrspace 3)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 48
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p3) :: (load (<4 x s32>) from unknown-address + 48, addrspace 3)
+ ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v16s32_align32
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 32, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<4 x s32>) from unknown-address + 16, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p3) :: (load (<4 x s32>) from unknown-address + 32, align 32, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 48
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p3) :: (load (<4 x s32>) from unknown-address + 48, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<16 x s32>) = G_LOAD %0 :: (load (<16 x s32>), align 32, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %1
@@ -13287,6 +15647,24 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s64_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 4, addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 8, align 4, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[LOAD]](s64), [[LOAD1]](s64)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s64_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -13876,6 +16254,90 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s64_align16
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
+ ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
+ ; GFX12-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p3) :: (load (s8) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p3) :: (load (s8) from unknown-address + 13, addrspace 3)
+ ; GFX12-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
+ ; GFX12-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p3) :: (load (s8) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p3) :: (load (s8) from unknown-address + 15, addrspace 3)
+ ; GFX12-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
+ ; GFX12-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
+ ; GFX12-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
+ ; GFX12-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
+ ; GFX12-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s64_align16
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 1, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -14020,6 +16482,32 @@ body: |
; GFX11-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v3s64_align32
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s64_align32
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
+ ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 3)
%2:_(<4 x s64>) = G_IMPLICIT_DEF
@@ -14154,6 +16642,28 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s64>) from unknown-address + 16, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v4s64_align32
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s64>) from unknown-address + 16, addrspace 3)
+ ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v4s64_align32
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p3) :: (load (<2 x s64>) from unknown-address + 16, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s64>) = G_LOAD %0 :: (load (<4 x s64>), align 32, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %1
@@ -14312,6 +16822,32 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2p1_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2p1_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -14402,6 +16938,22 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
; GFX11-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2p3_align8
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2p3_align8
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 8, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -14482,6 +17034,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_s32_from_1_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_s32_from_1_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s32) = G_LOAD %0 :: (load (s8), align 4, addrspace 3)
$vgpr0 = COPY %1
@@ -14562,6 +17128,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_s32_from_2_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_s32_from_2_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(s32) = G_LOAD %0 :: (load (s16), align 4, addrspace 3)
$vgpr0 = COPY %1
@@ -14653,6 +17233,22 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_s64_from_1_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_s64_from_1_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -14728,21 +17324,69 @@ body: |
; GFX10-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX10-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX11-LABEL: name: test_extload_local_s64_from_2_align4
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
- ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_extload_local_s64_from_2_align4
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; GFX11-TRUE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_extload_local_s64_from_2_align4
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -14833,6 +17477,22 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
; GFX11-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_s64_from_4_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_s64_from_4_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s32), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -14953,6 +17613,28 @@ body: |
; GFX11-UNALIGNED-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
; GFX11-UNALIGNED-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_s128_from_4_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX12-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_s128_from_4_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
%0:_(p3) = COPY $vgpr0
%1:_(s128) = G_LOAD %0 :: (load (s32), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -15028,21 +17710,69 @@ body: |
; GFX10-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX10-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX11-LABEL: name: test_extload_local_s64_from_2_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
- ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_extload_local_s64_from_2_align2
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; GFX11-TRUE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_extload_local_s64_from_2_align2
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -15133,6 +17863,22 @@ body: |
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_s64_from_1_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_s64_from_1_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s8), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s8), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -15213,6 +17959,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 1, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_v2s32_from_4_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 1, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_v2s32_from_4_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s16>), align 1, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -15293,6 +18053,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 2, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_v2s32_from_4_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 2, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_v2s32_from_4_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -15373,6 +18147,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_v2s32_from_4_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_v2s32_from_4_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s16>), addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s32>) = G_LOAD %0 :: (load (<2 x s16>), align 4, addrspace 3)
$vgpr0_vgpr1 = COPY %1
@@ -15453,6 +18241,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s16>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_v3s32_from_6_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s16>), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_v3s32_from_6_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s16>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s16>), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2 = COPY %1
@@ -15533,6 +18335,20 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 4, addrspace 3)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-LABEL: name: test_extload_local_v4s32_from_8_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 4, addrspace 3)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_extload_local_v4s32_from_8_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
%0:_(p3) = COPY $vgpr0
%1:_(<4 x s32>) = G_LOAD %0 :: (load (<4 x s16>), align 4, addrspace 3)
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
@@ -16314,6 +19130,123 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s96_align1
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s8) from unknown-address + 3, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s8) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p3) :: (load (s8) from unknown-address + 5, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s8) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s8) from unknown-address + 7, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s8) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p3) :: (load (s8) from unknown-address + 9, addrspace 3)
+ ; GFX12-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s8) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s8) from unknown-address + 11, addrspace 3)
+ ; GFX12-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
+ ; GFX12-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p3) :: (load (s8) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p3) :: (load (s8) from unknown-address + 13, addrspace 3)
+ ; GFX12-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
+ ; GFX12-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p3) :: (load (s8) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p3) :: (load (s8) from unknown-address + 15, addrspace 3)
+ ; GFX12-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
+ ; GFX12-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
+ ; GFX12-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p3) :: (load (s8) from unknown-address + 16, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p3) :: (load (s8) from unknown-address + 17, addrspace 3)
+ ; GFX12-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
+ ; GFX12-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p3) :: (load (s8) from unknown-address + 18, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p3) :: (load (s8) from unknown-address + 19, addrspace 3)
+ ; GFX12-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
+ ; GFX12-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
+ ; GFX12-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C1]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p3) :: (load (s8) from unknown-address + 20, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p3) :: (load (s8) from unknown-address + 21, addrspace 3)
+ ; GFX12-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
+ ; GFX12-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p3) :: (load (s8) from unknown-address + 22, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p3) :: (load (s8) from unknown-address + 23, addrspace 3)
+ ; GFX12-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
+ ; GFX12-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
+ ; GFX12-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
+ ; GFX12-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s96_align1
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 1, addrspace 3)
%2:_(s96) = G_EXTRACT %1, 0
@@ -16755,6 +19688,74 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s96_align2
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s16) from unknown-address + 2, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p3) :: (load (s16) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s16) from unknown-address + 6, addrspace 3)
+ ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
+ ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p3) :: (load (s16) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p3) :: (load (s16) from unknown-address + 10, addrspace 3)
+ ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p3) :: (load (s16) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p3) :: (load (s16) from unknown-address + 14, addrspace 3)
+ ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
+ ; GFX12-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p3) :: (load (s16) from unknown-address + 16, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p3) :: (load (s16) from unknown-address + 18, addrspace 3)
+ ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
+ ; GFX12-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s32)
+ ; GFX12-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p3) :: (load (s16) from unknown-address + 20, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p3) :: (load (s16) from unknown-address + 22, addrspace 3)
+ ; GFX12-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
+ ; GFX12-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
+ ; GFX12-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s96_align2
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 2, addrspace 3)
%2:_(s96) = G_EXTRACT %1, 0
@@ -17008,6 +20009,48 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s96_align4
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 4, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 8, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LOAD1]](s32), [[LOAD2]](s32)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD2]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p3) :: (load (s32) from unknown-address + 16, addrspace 3)
+ ; GFX12-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD2]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p3) :: (load (s32) from unknown-address + 20, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[LOAD3]](s32), [[LOAD4]](s32), [[LOAD5]](s32)
+ ; GFX12-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s96_align4
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 4, addrspace 3)
%2:_(s96) = G_EXTRACT %1, 0
@@ -17231,6 +20274,43 @@ body: |
; GFX11-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX11-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-LABEL: name: test_load_local_v2s96_align16
+ ; GFX12: liveins: $vgpr0
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s32) from unknown-address + 12, addrspace 3)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C1]](s32)
+ ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s32) from unknown-address + 16, addrspace 3)
+ ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s32)
+ ; GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p3) :: (load (s32) from unknown-address + 20, addrspace 3)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[LOAD1]](s32), [[LOAD2]](s32), [[LOAD3]](s32)
+ ; GFX12-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
+ ;
+ ; GFX12-UNALIGNED-LABEL: name: test_load_local_v2s96_align16
+ ; GFX12-UNALIGNED: liveins: $vgpr0
+ ; GFX12-UNALIGNED-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
+ ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 3)
+ ; GFX12-UNALIGNED-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
+ ; GFX12-UNALIGNED-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
+ ; GFX12-UNALIGNED-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
+ ; GFX12-UNALIGNED-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
%0:_(p3) = COPY $vgpr0
%1:_(<2 x s96>) = G_LOAD %0 :: (load (<2 x s96>), align 16, addrspace 3)
%2:_(s96) = G_EXTRACT %1, 0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir
index 6f8aaab1f878e..52dc8701e86f2 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir
@@ -4,16 +4,20 @@
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX10 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12,GFX12-FAKE16 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=tahiti -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=SI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=CI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX9 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX10 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX11 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX12 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX11,UNALIGNED_GFX11_TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX11,UNALIGNED_GFX11_FAKE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX12,UNALIGNED_GFX12_TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX12,UNALIGNED_GFX12_FAKE16 %s
---
name: test_load_private_s1_align1
@@ -452,19 +456,35 @@ body: |
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-LABEL: name: test_load_private_s16_align4
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_private_s16_align4
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_s16_align4
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_s16_align4
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_s16_align4
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_s16_align4
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_s16_align4
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -480,19 +500,35 @@ body: |
; UNALIGNED_GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
; UNALIGNED_GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_s16_align4
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_s16_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_s16_align4
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_s16_align4
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_s16_align4
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_s16_align4
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p5) = COPY $vgpr0
%1:_(s16) = G_LOAD %0 :: (load (s16), align 4, addrspace 5)
%2:_(s32) = G_ANYEXT %1
@@ -540,19 +576,35 @@ body: |
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-LABEL: name: test_load_private_s16_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_private_s16_align2
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_s16_align2
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -568,19 +620,35 @@ body: |
; UNALIGNED_GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; UNALIGNED_GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_s16_align2
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_s16_align2
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_s16_align2
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_s16_align2
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_s16_align2
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p5) = COPY $vgpr0
%1:_(s16) = G_LOAD %0 :: (load (s16), align 2, addrspace 5)
%2:_(s32) = G_ANYEXT %1
@@ -646,19 +714,35 @@ body: |
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-LABEL: name: test_load_private_s16_align1
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_private_s16_align1
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_s16_align1
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_s16_align1
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_s16_align1
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_s16_align1
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_s16_align1
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -7424,19 +7508,35 @@ body: |
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-LABEL: name: test_load_private_v2s8_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX12-LABEL: name: test_load_private_v2s8_align2
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_v2s8_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_v2s8_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_v2s8_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_v2s8_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_v2s8_align2
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -7452,19 +7552,35 @@ body: |
; UNALIGNED_GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
; UNALIGNED_GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_v2s8_align2
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_v2s8_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_v2s8_align2
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_v2s8_align2
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_v2s8_align2
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_v2s8_align2
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p5) = COPY $vgpr0
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 2, addrspace 5)
%2:_(s16) = G_BITCAST %1
@@ -7543,25 +7659,49 @@ body: |
; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX11-LABEL: name: test_load_private_v2s8_align1
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
- ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
- ;
- ; GFX12-LABEL: name: test_load_private_v2s8_align1
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
- ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_v2s8_align1
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[LOAD]], [[C]](s16)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_v2s8_align1
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_v2s8_align1
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[LOAD]], [[C]](s16)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_v2s8_align1
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[LOAD]](s32), [[LSHR]](s32)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_v2s8_align1
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -9347,31 +9487,53 @@ body: |
; UNALIGNED_GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; UNALIGNED_GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_v2s16_align2
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX11-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_v2s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_v2s16_align2
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_v2s16_align2
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_v2s16_align2
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_v2s16_align2
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(p5) = COPY $vgpr0
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 5)
$vgpr0 = COPY %1
@@ -9944,47 +10106,83 @@ body: |
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-LABEL: name: test_load_private_v3s16_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX12-LABEL: name: test_load_private_v3s16_align2
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_v3s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_v3s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_v3s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_v3s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_v3s16_align2
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -10028,47 +10226,83 @@ body: |
; UNALIGNED_GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_v3s16_align2
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; UNALIGNED_GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_v3s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX12-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; UNALIGNED_GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_v3s16_align2
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_v3s16_align2
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_v3s16_align2
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_v3s16_align2
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p5) = COPY $vgpr0
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 2, addrspace 5)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -10253,47 +10487,83 @@ body: |
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-LABEL: name: test_load_private_v3s16_align1
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, align 1, addrspace 5)
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, align 1, addrspace 5)
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX12-LABEL: name: test_load_private_v3s16_align1
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, align 1, addrspace 5)
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, align 1, addrspace 5)
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_v3s16_align1
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, align 1, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, align 1, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_v3s16_align1
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, align 1, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, align 1, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_v3s16_align1
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, align 1, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, align 1, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_v3s16_align1
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, align 1, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, align 1, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_v3s16_align1
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -10839,51 +11109,89 @@ body: |
; UNALIGNED_GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; UNALIGNED_GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_v4s16_align2
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s16) from unknown-address + 6, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; UNALIGNED_GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; UNALIGNED_GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; UNALIGNED_GFX11-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_v4s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s16) from unknown-address + 6, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; UNALIGNED_GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; UNALIGNED_GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; UNALIGNED_GFX12-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_v4s16_align2
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p5) :: (load (s16) from unknown-address + 6, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_v4s16_align2
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s16) from unknown-address + 6, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_v4s16_align2
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p5) :: (load (s16) from unknown-address + 6, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_v4s16_align2
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s16) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s16) from unknown-address + 4, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s16) from unknown-address + 6, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
%0:_(p5) = COPY $vgpr0
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 2, addrspace 5)
$vgpr0_vgpr1 = COPY %1
@@ -17034,21 +17342,37 @@ body: |
; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX11-LABEL: name: test_ext_load_private_s64_from_2_align4
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX12-LABEL: name: test_ext_load_private_s64_from_2_align4
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; GFX11-TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
; UNALIGNED_GFX9-LABEL: name: test_ext_load_private_s64_from_2_align4
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -17066,21 +17390,37 @@ body: |
; UNALIGNED_GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; UNALIGNED_GFX11-LABEL: name: test_ext_load_private_s64_from_2_align4
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_ext_load_private_s64_from_2_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align4
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p5) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 5)
$vgpr0_vgpr1 = COPY %1
@@ -17361,21 +17701,37 @@ body: |
; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX11-LABEL: name: test_ext_load_private_s64_from_2_align2
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX12-LABEL: name: test_ext_load_private_s64_from_2_align2
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; GFX11-TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
; UNALIGNED_GFX9-LABEL: name: test_ext_load_private_s64_from_2_align2
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -17393,21 +17749,37 @@ body: |
; UNALIGNED_GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; UNALIGNED_GFX11-LABEL: name: test_ext_load_private_s64_from_2_align2
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX11-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_ext_load_private_s64_from_2_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_ext_load_private_s64_from_2_align2
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), align 4, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p5) = COPY $vgpr0
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 5)
$vgpr0_vgpr1 = COPY %1
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
index 00440eeb27d15..423e60b5dc5b5 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
@@ -3,9 +3,11 @@
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=CI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX9 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX9 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX10 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX12,GFX12-FAKE16 %s
---
name: test_store_global_s1_align1
@@ -52,6 +54,36 @@ body: |
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[AND]], [[C]]
; GFX9-NEXT: G_STORE [[AND1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s1_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX10-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[AND]], [[C]]
+ ; GFX10-NEXT: G_STORE [[AND1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s1_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[AND]], [[C]]
+ ; GFX11-NEXT: G_STORE [[AND1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s1_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[AND]], [[C]]
+ ; GFX12-NEXT: G_STORE [[AND1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s1) = G_TRUNC %1
@@ -99,6 +131,33 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 127
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
; GFX9-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s7_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 127
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX10-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s7_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 127
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX11-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s7_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 127
+ ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX12-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s7) = G_TRUNC %1
@@ -138,6 +197,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s8_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s8_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s8_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s8) = G_TRUNC %1
@@ -191,6 +271,43 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s16_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_store_global_s16_align1
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
+ ; GFX11-TRUE16-NEXT: G_STORE [[TRUNC]](s16), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_store_global_s16_align1
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-FAKE16-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_store_global_s16_align1
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
+ ; GFX12-TRUE16-NEXT: G_STORE [[TRUNC]](s16), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_store_global_s16_align1
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-FAKE16-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s16) = G_TRUNC %1
@@ -230,6 +347,43 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s16_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_store_global_s16_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
+ ; GFX11-TRUE16-NEXT: G_STORE [[TRUNC]](s16), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_store_global_s16_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-FAKE16-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_store_global_s16_align2
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
+ ; GFX12-TRUE16-NEXT: G_STORE [[TRUNC]](s16), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_store_global_s16_align2
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-FAKE16-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s16) = G_TRUNC %1
@@ -269,6 +423,43 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_store_global_s16_align4
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
+ ; GFX11-TRUE16-NEXT: G_STORE [[TRUNC]](s16), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_store_global_s16_align4
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-FAKE16-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_store_global_s16_align4
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
+ ; GFX12-TRUE16-NEXT: G_STORE [[TRUNC]](s16), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_store_global_s16_align4
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-FAKE16-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s16) = G_TRUNC %1
@@ -332,6 +523,45 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
; GFX9-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s24_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ; GFX10-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s24_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ; GFX11-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s24_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
+ ; GFX12-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s24) = G_TRUNC %1
@@ -395,6 +625,45 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
; GFX9-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s24_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ; GFX10-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s24_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ; GFX11-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s24_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
+ ; GFX12-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s24) = G_TRUNC %1
@@ -472,6 +741,45 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
; GFX9-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s24_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ; GFX10-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s24_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ; GFX11-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s24_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
+ ; GFX12-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s24) = G_TRUNC %1
@@ -519,6 +827,33 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 33554431
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
; GFX9-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s25_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 33554431
+ ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX10-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s25_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 33554431
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX11-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s25_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 33554431
+ ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+ ; GFX12-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s25) = G_TRUNC %1
@@ -617,6 +952,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s32_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
G_STORE %1, %0 :: (store (s32), align 1, addrspace 1)
@@ -667,6 +1023,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s32_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
G_STORE %1, %0 :: (store (s32), align 2, addrspace 1)
@@ -705,6 +1082,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX11-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s32_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+ ; GFX12-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
G_STORE %1, %0 :: (store (s32), align 4, addrspace 1)
@@ -780,6 +1178,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p3_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p3_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX11-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p3_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX12-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p3) = COPY $vgpr2
G_STORE %1, %0 :: (store (p3), align 1, addrspace 1)
@@ -832,6 +1251,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p3_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p3_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX11-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p3_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX12-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p3) = COPY $vgpr2
G_STORE %1, %0 :: (store (p3), align 2, addrspace 1)
@@ -870,6 +1310,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p3_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX10-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p3_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX11-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p3_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
+ ; GFX12-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p3) = COPY $vgpr2
G_STORE %1, %0 :: (store (p3), align 4, addrspace 1)
@@ -985,6 +1446,81 @@ body: |
; GFX9-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
; GFX9-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s48_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX10-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+ ; GFX10-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_store_global_s48_align1
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-TRUE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX11-TRUE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s64)
+ ; GFX11-TRUE16-NEXT: G_STORE [[TRUNC1]](s16), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_store_global_s48_align1
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-FAKE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX11-FAKE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+ ; GFX11-FAKE16-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_store_global_s48_align1
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-TRUE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX12-TRUE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s64)
+ ; GFX12-TRUE16-NEXT: G_STORE [[TRUNC1]](s16), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_store_global_s48_align1
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-FAKE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX12-FAKE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+ ; GFX12-FAKE16-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
%2:_(s48) = G_TRUNC %1
@@ -1068,6 +1604,81 @@ body: |
; GFX9-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
; GFX9-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s48_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX10-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+ ; GFX10-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_store_global_s48_align2
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-TRUE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX11-TRUE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s64)
+ ; GFX11-TRUE16-NEXT: G_STORE [[TRUNC1]](s16), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_store_global_s48_align2
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-FAKE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX11-FAKE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+ ; GFX11-FAKE16-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_store_global_s48_align2
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-TRUE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX12-TRUE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s64)
+ ; GFX12-TRUE16-NEXT: G_STORE [[TRUNC1]](s16), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_store_global_s48_align2
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-FAKE16-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
+ ; GFX12-FAKE16-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
+ ; GFX12-FAKE16-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
%2:_(s48) = G_TRUNC %1
@@ -1210,6 +1821,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s64_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s64_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s64_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (s64), align 1, addrspace 1)
@@ -1284,6 +1916,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s64_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s64_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s64_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (s64), align 2, addrspace 1)
@@ -1322,6 +1975,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s64_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s64_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s64_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (s64), align 4, addrspace 1)
@@ -1360,6 +2034,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s64_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s64_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s64_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (s64), align 8, addrspace 1)
@@ -1398,6 +2093,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s64_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s64_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s64_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (s64), align 16, addrspace 1)
@@ -1516,6 +2232,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p0_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p0_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p0_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p0) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p0), align 1, addrspace 1)
@@ -1592,6 +2329,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p0_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p0_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p0_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p0) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p0), align 2, addrspace 1)
@@ -1630,6 +2388,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p0_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p0_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p0_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p0) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p0), align 4, addrspace 1)
@@ -1668,6 +2447,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p0_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p0_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p0_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p0) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p0), align 8, addrspace 1)
@@ -1706,6 +2506,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p0_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p0_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p0_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p0) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p0), align 16, addrspace 1)
@@ -1824,6 +2645,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p999_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p999_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p999_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p999) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p999), align 1, addrspace 1)
@@ -1900,6 +2742,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p999_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p999_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p999_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p999) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p999), align 2, addrspace 1)
@@ -1938,6 +2801,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p999_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p999_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p999_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p999) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p999), align 4, addrspace 1)
@@ -1976,6 +2860,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p999_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p999_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p999_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p999) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p999), align 8, addrspace 1)
@@ -2014,6 +2919,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_p999_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_p999_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_p999_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(p999) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (p999), align 16, addrspace 1)
@@ -2122,6 +3048,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s32_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x s32>), align 1, addrspace 1)
@@ -2188,6 +3135,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s32_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x s32>), align 2, addrspace 1)
@@ -2226,6 +3194,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s32_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x s32>), align 4, addrspace 1)
@@ -2264,6 +3253,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s32_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x s32>), align 8, addrspace 1)
@@ -2302,6 +3312,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s32_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x s32>), align 16, addrspace 1)
@@ -2414,6 +3445,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p3_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p3_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p3_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x p3>), align 1, addrspace 1)
@@ -2484,6 +3539,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p3_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p3_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p3_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x p3>), align 2, addrspace 1)
@@ -2526,6 +3605,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p3_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p3_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p3_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x p3>), align 4, addrspace 1)
@@ -2568,6 +3671,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p3_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p3_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p3_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x p3>), align 8, addrspace 1)
@@ -2610,6 +3737,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p3_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p3_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p3_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p3>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<2 x p3>), align 16, addrspace 1)
@@ -2722,6 +3873,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s16_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s16_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s16_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<4 x s16>), align 1, addrspace 1)
@@ -2790,6 +3962,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s16_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s16_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s16_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<4 x s16>), align 2, addrspace 1)
@@ -2828,6 +4021,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s16_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s16_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<4 x s16>), align 4, addrspace 1)
@@ -2866,6 +4080,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s16_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s16_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s16_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<4 x s16>), align 8, addrspace 1)
@@ -2904,6 +4139,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s16_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s16_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s16_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store (<4 x s16>), align 16, addrspace 1)
@@ -3045,6 +4301,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v3s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v3s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v3s32_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (<3 x s32>), align 1, addrspace 1)
@@ -3125,6 +4402,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v3s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v3s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v3s32_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (<3 x s32>), align 2, addrspace 1)
@@ -3168,6 +4466,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v3s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v3s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v3s32_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (<3 x s32>), align 4, addrspace 1)
@@ -3211,6 +4530,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v3s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v3s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v3s32_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (<3 x s32>), align 8, addrspace 1)
@@ -3254,6 +4594,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v3s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v3s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v3s32_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (<3 x s32>), align 16, addrspace 1)
@@ -3428,6 +4789,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s32_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<4 x s32>), align 1, addrspace 1)
@@ -3522,6 +4904,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s32_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<4 x s32>), align 2, addrspace 1)
@@ -3560,6 +4963,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s32_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<4 x s32>), align 4, addrspace 1)
@@ -3598,6 +5022,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s32_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<4 x s32>), align 8, addrspace 1)
@@ -3636,6 +5081,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v4s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v4s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v4s32_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<4 x s32>), align 16, addrspace 1)
@@ -3828,6 +5294,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s64_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s64_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s64_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x s64>), align 1, addrspace 1)
@@ -3940,6 +5427,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s64_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s64_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s64_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x s64>), align 2, addrspace 1)
@@ -3978,6 +5486,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s64_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s64_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s64_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x s64>), align 4, addrspace 1)
@@ -4016,6 +5545,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s64_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s64_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s64_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x s64>), align 8, addrspace 1)
@@ -4054,6 +5604,27 @@ body: |
; GFX9-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s64_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s64_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s64_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x s64>), align 16, addrspace 1)
@@ -4232,6 +5803,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s16_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s16_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s16_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 1, addrspace 1)
@@ -4330,6 +5925,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s16_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s16_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s16_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 2, addrspace 1)
@@ -4372,6 +5991,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s16_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s16_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 4, addrspace 1)
@@ -4414,6 +6057,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s16_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s16_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s16_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 8, addrspace 1)
@@ -4456,6 +6123,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s16_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s16_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s16_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<8 x s16>), align 16, addrspace 1)
@@ -4634,6 +6325,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p0_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p0_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p0_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 1, addrspace 1)
@@ -4732,6 +6447,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p0_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p0_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p0_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 2, addrspace 1)
@@ -4774,6 +6513,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p0_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p0_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p0_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 4, addrspace 1)
@@ -4816,6 +6579,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p0_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p0_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p0_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 8, addrspace 1)
@@ -4858,6 +6645,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2p0_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2p0_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2p0_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (<2 x p0>), align 16, addrspace 1)
@@ -5003,6 +6814,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s96_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s96_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s96_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 1, addrspace 1)
@@ -5087,6 +6922,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s96_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s96_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s96_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 2, addrspace 1)
@@ -5134,6 +6993,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s96_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s96_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s96_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 4, addrspace 1)
@@ -5181,6 +7064,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s96_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s96_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s96_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 8, addrspace 1)
@@ -5228,6 +7135,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s96_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s96_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s96_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s96) = COPY $vgpr2_vgpr3_vgpr4
G_STORE %1, %0 :: (store (s96), align 16, addrspace 1)
@@ -5406,6 +7337,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s128_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s128_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s128_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 1, addrspace 1)
@@ -5504,6 +7459,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s128_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s128_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s128_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 2, addrspace 1)
@@ -5546,6 +7525,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s128_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s128_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s128_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 4, addrspace 1)
@@ -5588,6 +7591,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s128_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s128_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s128_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 8, addrspace 1)
@@ -5630,6 +7657,30 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s128_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s128_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX11-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s128_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
+ ; GFX12-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
G_STORE %1, %0 :: (store (s128), align 16, addrspace 1)
@@ -5847,6 +7898,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5s32_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x s32>), align 1, addrspace 1)
@@ -5965,6 +8052,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5s32_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x s32>), align 2, addrspace 1)
@@ -6023,6 +8146,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5s32_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x s32>), align 4, addrspace 1)
@@ -6081,6 +8240,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5s32_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x s32>), align 8, addrspace 1)
@@ -6139,6 +8334,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5s32_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x s32>), align 16, addrspace 1)
@@ -6359,6 +8590,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5p3_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5p3_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5p3_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x p3>), align 1, addrspace 1)
@@ -6481,6 +8751,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5p3_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5p3_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5p3_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x p3>), align 2, addrspace 1)
@@ -6543,6 +8852,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5p3_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5p3_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5p3_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x p3>), align 4, addrspace 1)
@@ -6605,6 +8953,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5p3_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5p3_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5p3_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x p3>), align 8, addrspace 1)
@@ -6667,6 +9054,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v5p3_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v5p3_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v5p3_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (<5 x p3>), align 16, addrspace 1)
@@ -6729,6 +9155,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v10s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v10s16_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v10s16_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<10 x s16>) = G_IMPLICIT_DEF
G_STORE %1, %0 :: (store (<10 x s16>), align 16, addrspace 1)
@@ -6858,6 +9323,119 @@ body: |
; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
; GFX9-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v11s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+ ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
+ ; GFX10-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+ ;
+ ; GFX11-TRUE16-LABEL: name: test_store_global_v11s16_align4
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX11-TRUE16-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-TRUE16-NEXT: G_STORE [[DEF]](s16), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
+ ; GFX11-TRUE16-NEXT: G_STORE [[DEF]](s16), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
+ ; GFX11-TRUE16-NEXT: G_STORE [[DEF]](s16), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_store_global_v11s16_align4
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX11-FAKE16-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-FAKE16-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
+ ; GFX11-FAKE16-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
+ ; GFX11-FAKE16-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_store_global_v11s16_align4
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX12-TRUE16-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-TRUE16-NEXT: G_STORE [[DEF]](s16), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
+ ; GFX12-TRUE16-NEXT: G_STORE [[DEF]](s16), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
+ ; GFX12-TRUE16-NEXT: G_STORE [[DEF]](s16), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_store_global_v11s16_align4
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+ ; GFX12-FAKE16-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-FAKE16-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
+ ; GFX12-FAKE16-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
+ ; GFX12-FAKE16-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<11 x s16>) = G_IMPLICIT_DEF
G_STORE %1, %0 :: (store (<11 x s16>), align 16, addrspace 1)
@@ -6920,6 +9498,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV2]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v12s16_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>), [[UV2:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
+ ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[UV]](<2 x s32>), [[UV1]](<2 x s32>)
+ ; GFX10-NEXT: G_STORE [[CONCAT_VECTORS]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV2]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v12s16_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>), [[UV2:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
+ ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[UV]](<2 x s32>), [[UV1]](<2 x s32>)
+ ; GFX11-NEXT: G_STORE [[CONCAT_VECTORS]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV2]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v12s16_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>), [[UV2:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
+ ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[UV]](<2 x s32>), [[UV1]](<2 x s32>)
+ ; GFX12-NEXT: G_STORE [[CONCAT_VECTORS]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV2]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<12 x s16>) = G_IMPLICIT_DEF
G_STORE %1, %0 :: (store (<12 x s16>), align 16, addrspace 1)
@@ -7141,6 +9758,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s160_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s160_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s160_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (s160), align 1, addrspace 1)
@@ -7263,6 +9919,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s160_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s160_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s160_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (s160), align 2, addrspace 1)
@@ -7325,6 +10020,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s160_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s160_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s160_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (s160), align 4, addrspace 1)
@@ -7387,6 +10121,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s160_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s160_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s160_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (s160), align 8, addrspace 1)
@@ -7449,6 +10222,45 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s160_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s160_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s160_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
G_STORE %1, %0 :: (store (s160), align 16, addrspace 1)
@@ -7759,6 +10571,39 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s32_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (<8 x s32>), align 1, addrspace 1)
@@ -7913,6 +10758,39 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s32_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (<8 x s32>), align 2, addrspace 1)
@@ -7967,6 +10845,39 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s32_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (<8 x s32>), align 4, addrspace 1)
@@ -8021,6 +10932,39 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s32_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (<8 x s32>), align 8, addrspace 1)
@@ -8075,6 +11019,39 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s32_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (<8 x s32>), align 16, addrspace 1)
@@ -8133,6 +11110,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v2s128_align32
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s128>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](<2 x s128>)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v2s128_align32
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s128>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](<2 x s128>)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v2s128_align32
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s128>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](<2 x s128>)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s128>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (<2 x s128>), align 32, addrspace 1)
@@ -8447,6 +11460,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s256_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s256_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s256_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (s256), align 1, addrspace 1)
@@ -8605,6 +11654,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s256_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s256_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s256_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (s256), align 2, addrspace 1)
@@ -8663,6 +11748,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s256_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s256_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s256_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (s256), align 4, addrspace 1)
@@ -8721,6 +11842,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s256_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s256_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s256_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (s256), align 8, addrspace 1)
@@ -8779,6 +11936,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s256_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s256_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s256_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (s256), align 16, addrspace 1)
@@ -8837,6 +12030,42 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_s256_align32
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_s256_align32
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_s256_align32
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (s256), align 32, addrspace 1)
@@ -8891,6 +12120,39 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v8s32_align32
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v8s32_align32
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX11-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v8s32_align32
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
+ ; GFX12-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
G_STORE %1, %0 :: (store (<8 x s32>), align 32, addrspace 1)
@@ -9258,6 +12520,66 @@ body: |
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v9s32_align1
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v9s32_align1
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX11-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX11-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v9s32_align1
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX12-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX12-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
%2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
@@ -9453,6 +12775,66 @@ body: |
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v9s32_align2
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v9s32_align2
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX11-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX11-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v9s32_align2
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX12-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX12-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
%2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
@@ -9546,6 +12928,66 @@ body: |
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v9s32_align4
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v9s32_align4
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX11-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX11-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v9s32_align4
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX12-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX12-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
%2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
@@ -9639,6 +13081,66 @@ body: |
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v9s32_align8
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v9s32_align8
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX11-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX11-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v9s32_align8
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX12-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX12-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
%2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
@@ -9732,6 +13234,66 @@ body: |
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
+ ;
+ ; GFX10-LABEL: name: test_store_global_v9s32_align16
+ ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX10-NEXT: {{ $}}
+ ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
+ ;
+ ; GFX11-LABEL: name: test_store_global_v9s32_align16
+ ; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX11-NEXT: {{ $}}
+ ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX11-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX11-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX11-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX11-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
+ ;
+ ; GFX12-LABEL: name: test_store_global_v9s32_align16
+ ; GFX12: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
+ ; GFX12-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+ ; GFX12-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+ ; GFX12-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+ ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+ ; GFX12-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+ ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+ ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+ ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
+ ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
%2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll
index 10e83b70a57d4..ec4f3efa2d150 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll
@@ -239,47 +239,53 @@ define amdgpu_kernel void @store_lds_v4i32_align1(ptr addrspace(3) %out, <4 x i3
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_and_b32 s6, 0xffff, s0
; GFX11-NEXT: s_lshr_b32 s5, s0, 16
-; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s4
+; GFX11-NEXT: v_mov_b16_e32 v0.l, s0
; GFX11-NEXT: s_lshr_b32 s0, s0, 24
-; GFX11-NEXT: s_lshr_b32 s4, s1, 16
; GFX11-NEXT: s_and_b32 s7, 0xffff, s1
; GFX11-NEXT: s_lshr_b32 s6, s6, 8
-; GFX11-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v3, s2
-; GFX11-NEXT: s_lshr_b32 s1, s1, 24
-; GFX11-NEXT: v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s0
+; GFX11-NEXT: v_mov_b32_e32 v5, s4
+; GFX11-NEXT: v_mov_b16_e32 v2.l, s0
; GFX11-NEXT: s_lshr_b32 s0, s7, 8
-; GFX11-NEXT: v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s6
+; GFX11-NEXT: v_mov_b16_e32 v3.l, s6
+; GFX11-NEXT: v_mov_b16_e32 v1.h, s5
+; GFX11-NEXT: s_lshr_b32 s4, s1, 16
+; GFX11-NEXT: v_mov_b16_e32 v0.h, s1
+; GFX11-NEXT: s_lshr_b32 s1, s1, 24
; GFX11-NEXT: s_and_b32 s9, 0xffff, s2
-; GFX11-NEXT: v_dual_mov_b32 v8, s1 :: v_dual_mov_b32 v9, s0
+; GFX11-NEXT: v_mov_b16_e32 v4.l, s0
; GFX11-NEXT: s_lshr_b32 s0, s2, 24
+; GFX11-NEXT: v_mov_b16_e32 v2.h, s4
+; GFX11-NEXT: v_mov_b16_e32 v3.h, s1
; GFX11-NEXT: s_lshr_b32 s1, s9, 8
-; GFX11-NEXT: ds_store_b8 v1, v0
-; GFX11-NEXT: ds_store_b8 v1, v7 offset:1
-; GFX11-NEXT: ds_store_b8 v1, v4 offset:2
-; GFX11-NEXT: ds_store_b8 v1, v5 offset:3
-; GFX11-NEXT: ds_store_b8 v1, v2 offset:4
-; GFX11-NEXT: ds_store_b8 v1, v9 offset:5
-; GFX11-NEXT: ds_store_b8 v1, v6 offset:6
-; GFX11-NEXT: ds_store_b8 v1, v8 offset:7
-; GFX11-NEXT: v_mov_b32_e32 v4, s0
+; GFX11-NEXT: ds_store_b8 v5, v0
+; GFX11-NEXT: ds_store_b8 v5, v3 offset:1
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v1 offset:2
+; GFX11-NEXT: ds_store_b8 v5, v2 offset:3
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v0 offset:4
+; GFX11-NEXT: ds_store_b8 v5, v4 offset:5
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v2 offset:6
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v3 offset:7
+; GFX11-NEXT: v_mov_b16_e32 v1.h, s0
; GFX11-NEXT: s_and_b32 s0, 0xffff, s3
; GFX11-NEXT: s_lshr_b32 s8, s2, 16
-; GFX11-NEXT: v_dual_mov_b32 v0, s1 :: v_dual_mov_b32 v5, s3
+; GFX11-NEXT: v_mov_b16_e32 v1.l, s2
+; GFX11-NEXT: v_mov_b16_e32 v0.l, s1
; GFX11-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-NEXT: v_mov_b16_e32 v0.h, s8
; GFX11-NEXT: s_lshr_b32 s1, s3, 16
-; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT: v_dual_mov_b32 v2, s8 :: v_dual_mov_b32 v7, s1
-; GFX11-NEXT: v_mov_b32_e32 v6, s0
+; GFX11-NEXT: v_mov_b16_e32 v2.h, s0
; GFX11-NEXT: s_lshr_b32 s0, s3, 24
-; GFX11-NEXT: v_mov_b32_e32 v8, s0
-; GFX11-NEXT: ds_store_b8 v1, v3 offset:8
-; GFX11-NEXT: ds_store_b8 v1, v0 offset:9
-; GFX11-NEXT: ds_store_b8 v1, v2 offset:10
-; GFX11-NEXT: ds_store_b8 v1, v4 offset:11
-; GFX11-NEXT: ds_store_b8 v1, v5 offset:12
-; GFX11-NEXT: ds_store_b8 v1, v6 offset:13
-; GFX11-NEXT: ds_store_b8 v1, v7 offset:14
-; GFX11-NEXT: ds_store_b8 v1, v8 offset:15
+; GFX11-NEXT: v_mov_b16_e32 v2.l, s3
+; GFX11-NEXT: v_mov_b16_e32 v3.l, s1
+; GFX11-NEXT: v_mov_b16_e32 v3.h, s0
+; GFX11-NEXT: ds_store_b8 v5, v1 offset:8
+; GFX11-NEXT: ds_store_b8 v5, v0 offset:9
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v0 offset:10
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v1 offset:11
+; GFX11-NEXT: ds_store_b8 v5, v2 offset:12
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v2 offset:13
+; GFX11-NEXT: ds_store_b8 v5, v3 offset:14
+; GFX11-NEXT: ds_store_b8_d16_hi v5, v3 offset:15
; GFX11-NEXT: s_endpgm
store <4 x i32> %x, ptr addrspace(3) %out, align 1
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.96.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.96.ll
index db5d0fdb5c036..158b5d7450609 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.96.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.96.ll
@@ -206,37 +206,43 @@ define amdgpu_kernel void @store_lds_v3i32_align1(ptr addrspace(3) %out, <3 x i3
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: s_load_b32 s3, s[4:5], 0x0
; GFX11-NEXT: s_and_b32 s5, 0xffff, s0
-; GFX11-NEXT: s_lshr_b32 s7, s2, 16
-; GFX11-NEXT: s_lshr_b32 s5, s5, 8
; GFX11-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-NEXT: s_lshr_b32 s5, s5, 8
+; GFX11-NEXT: v_mov_b16_e32 v0.l, s0
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s3
+; GFX11-NEXT: v_mov_b32_e32 v6, s3
; GFX11-NEXT: s_lshr_b32 s0, s0, 24
; GFX11-NEXT: s_lshr_b32 s3, s1, 16
; GFX11-NEXT: s_and_b32 s6, 0xffff, s1
-; GFX11-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v3, s2
+; GFX11-NEXT: v_mov_b16_e32 v0.h, s1
; GFX11-NEXT: s_lshr_b32 s1, s1, 24
; GFX11-NEXT: s_and_b32 s8, 0xffff, s2
-; GFX11-NEXT: v_dual_mov_b32 v8, s7 :: v_dual_mov_b32 v9, s5
-; GFX11-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_mov_b32 v5, s0
+; GFX11-NEXT: v_mov_b16_e32 v4.l, s5
+; GFX11-NEXT: v_mov_b16_e32 v1.h, s4
+; GFX11-NEXT: s_lshr_b32 s7, s2, 16
+; GFX11-NEXT: v_mov_b16_e32 v1.l, s2
; GFX11-NEXT: s_lshr_b32 s2, s2, 24
+; GFX11-NEXT: v_mov_b16_e32 v2.l, s0
; GFX11-NEXT: s_lshr_b32 s0, s6, 8
-; GFX11-NEXT: v_dual_mov_b32 v6, s3 :: v_dual_mov_b32 v7, s1
+; GFX11-NEXT: v_mov_b16_e32 v3.l, s1
; GFX11-NEXT: s_lshr_b32 s1, s8, 8
-; GFX11-NEXT: v_dual_mov_b32 v10, s2 :: v_dual_mov_b32 v11, s0
-; GFX11-NEXT: v_mov_b32_e32 v12, s1
-; GFX11-NEXT: ds_store_b8 v1, v0
-; GFX11-NEXT: ds_store_b8 v1, v9 offset:1
-; GFX11-NEXT: ds_store_b8 v1, v4 offset:2
-; GFX11-NEXT: ds_store_b8 v1, v5 offset:3
-; GFX11-NEXT: ds_store_b8 v1, v2 offset:4
-; GFX11-NEXT: ds_store_b8 v1, v11 offset:5
-; GFX11-NEXT: ds_store_b8 v1, v6 offset:6
-; GFX11-NEXT: ds_store_b8 v1, v7 offset:7
-; GFX11-NEXT: ds_store_b8 v1, v3 offset:8
-; GFX11-NEXT: ds_store_b8 v1, v12 offset:9
-; GFX11-NEXT: ds_store_b8 v1, v8 offset:10
-; GFX11-NEXT: ds_store_b8 v1, v10 offset:11
+; GFX11-NEXT: v_mov_b16_e32 v2.h, s3
+; GFX11-NEXT: v_mov_b16_e32 v3.h, s7
+; GFX11-NEXT: v_mov_b16_e32 v4.h, s2
+; GFX11-NEXT: v_mov_b16_e32 v5.l, s0
+; GFX11-NEXT: v_mov_b16_e32 v5.h, s1
+; GFX11-NEXT: ds_store_b8 v6, v0
+; GFX11-NEXT: ds_store_b8 v6, v4 offset:1
+; GFX11-NEXT: ds_store_b8_d16_hi v6, v1 offset:2
+; GFX11-NEXT: ds_store_b8 v6, v2 offset:3
+; GFX11-NEXT: ds_store_b8_d16_hi v6, v0 offset:4
+; GFX11-NEXT: ds_store_b8 v6, v5 offset:5
+; GFX11-NEXT: ds_store_b8_d16_hi v6, v2 offset:6
+; GFX11-NEXT: ds_store_b8 v6, v3 offset:7
+; GFX11-NEXT: ds_store_b8 v6, v1 offset:8
+; GFX11-NEXT: ds_store_b8_d16_hi v6, v5 offset:9
+; GFX11-NEXT: ds_store_b8_d16_hi v6, v3 offset:10
+; GFX11-NEXT: ds_store_b8_d16_hi v6, v4 offset:11
; GFX11-NEXT: s_endpgm
store <3 x i32> %x, ptr addrspace(3) %out, align 1
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll
index a30e3115cc463..49cc38a0af276 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll
@@ -340,8 +340,11 @@ define i16 @v_uaddsat_v2i8(i16 %lhs.arg, i16 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, v1 clamp
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v0 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
-; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_uaddsat_v2i8:
@@ -442,86 +445,33 @@ define amdgpu_ps i16 @s_uaddsat_v2i8(i16 inreg %lhs.arg, i16 inreg %rhs.arg) {
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
-; GFX10-LABEL: s_uaddsat_v2i8:
-; GFX10: ; %bb.0:
-; GFX10-NEXT: s_lshr_b32 s2, s0, 8
-; GFX10-NEXT: s_lshr_b32 s3, s1, 8
-; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10-NEXT: s_lshr_b32 s2, s0, 16
-; GFX10-NEXT: s_lshr_b32 s3, s1, 16
-; GFX10-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX10-NEXT: s_lshl_b32 s2, s2, 8
-; GFX10-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX10-NEXT: s_lshl_b32 s3, s3, 8
-; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10-NEXT: v_pk_add_u16 v0, s0, s1 clamp
-; GFX10-NEXT: v_readfirstlane_b32 s0, v0
-; GFX10-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX10-NEXT: s_lshr_b32 s0, s0, 16
-; GFX10-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX10-NEXT: s_lshr_b32 s0, s0, 8
-; GFX10-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX10-NEXT: s_lshr_b32 s1, s0, 16
-; GFX10-NEXT: s_and_b32 s0, s0, 0xff
-; GFX10-NEXT: s_and_b32 s1, s1, 0xff
-; GFX10-NEXT: s_lshl_b32 s1, s1, 8
-; GFX10-NEXT: s_or_b32 s0, s0, s1
-; GFX10-NEXT: ; return to shader part epilog
-;
-; GFX11-TRUE16-LABEL: s_uaddsat_v2i8:
-; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, s0, s1 clamp
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-TRUE16-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s0, 8
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-TRUE16-NEXT: ; return to shader part epilog
-;
-; GFX11-FAKE16-LABEL: s_uaddsat_v2i8:
-; GFX11-FAKE16: ; %bb.0:
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v0, s0, s1 clamp
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-FAKE16-NEXT: ; return to shader part epilog
+; GFX10PLUS-LABEL: s_uaddsat_v2i8:
+; GFX10PLUS: ; %bb.0:
+; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 8
+; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 8
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 16
+; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 16
+; GFX10PLUS-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX10PLUS-NEXT: s_lshl_b32 s2, s2, 8
+; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX10PLUS-NEXT: s_lshl_b32 s3, s3, 8
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10PLUS-NEXT: v_pk_add_u16 v0, s0, s1 clamp
+; GFX10PLUS-NEXT: v_readfirstlane_b32 s0, v0
+; GFX10PLUS-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 16
+; GFX10PLUS-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 8
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX10PLUS-NEXT: s_lshr_b32 s1, s0, 16
+; GFX10PLUS-NEXT: s_and_b32 s0, s0, 0xff
+; GFX10PLUS-NEXT: s_and_b32 s1, s1, 0xff
+; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 8
+; GFX10PLUS-NEXT: s_or_b32 s0, s0, s1
+; GFX10PLUS-NEXT: ; return to shader part epilog
%lhs = bitcast i16 %lhs.arg to <2 x i8>
%rhs = bitcast i16 %rhs.arg to <2 x i8>
%result = call <2 x i8> @llvm.uadd.sat.v2i8(<2 x i8> %lhs, <2 x i8> %rhs)
@@ -679,21 +629,19 @@ define i32 @v_uaddsat_v4i8(i32 %lhs.arg, i32 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v0, 8, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v2, 8, v3 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v3, 8, v6 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, v1 clamp
-; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v6 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v2, v3 clamp
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v0, 8, v0 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v2, v1 clamp
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xff, v2
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v1.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 8
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v1, v1, 16, 8
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v3
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 24, v1
; GFX11-TRUE16-NEXT: v_and_or_b32 v0, 0xff, v0, v2
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 24, v3
-; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v2
+; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v3, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_uaddsat_v4i8:
@@ -928,107 +876,57 @@ define amdgpu_ps i32 @s_uaddsat_v4i8(i32 inreg %lhs.arg, i32 inreg %rhs.arg) {
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
-; GFX11-TRUE16-LABEL: s_uaddsat_v4i8:
-; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s0, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 24
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
-; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s2, 16
-; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s3, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, s2, s3 clamp
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, s0, s1 clamp
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s1, v1
-; GFX11-TRUE16-NEXT: s_and_b32 s2, s0, 0xffff
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
-; GFX11-TRUE16-NEXT: s_and_b32 s2, s1, 0xffff
-; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
-; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s2, s1
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s0, 8
-; GFX11-TRUE16-NEXT: s_and_b32 s1, s1, 0xff
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s1, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 24
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-TRUE16-NEXT: ; return to shader part epilog
-;
-; GFX11-FAKE16-LABEL: s_uaddsat_v4i8:
-; GFX11-FAKE16: ; %bb.0:
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s0, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 24
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
-; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s2, 16
-; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s3, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v0, s2, s3 clamp
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v1, s0, s1 clamp
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s1, v1
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s0, 0xffff
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xffff
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s3, 0x80008
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s3, s1
-; GFX11-FAKE16-NEXT: s_bfe_u32 s2, s0, 0x80010
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s3, 16
-; GFX11-FAKE16-NEXT: s_bfe_u32 s1, s1, 0x80010
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 24
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-FAKE16-NEXT: ; return to shader part epilog
+; GFX11-LABEL: s_uaddsat_v4i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-NEXT: s_lshr_b32 s4, s1, 8
+; GFX11-NEXT: s_lshr_b32 s5, s1, 24
+; GFX11-NEXT: s_pack_ll_b32_b16 s2, s0, s2
+; GFX11-NEXT: s_pack_hl_b32_b16 s0, s0, s3
+; GFX11-NEXT: s_pack_ll_b32_b16 s3, s1, s4
+; GFX11-NEXT: s_lshr_b32 s4, s2, 16
+; GFX11-NEXT: s_pack_hl_b32_b16 s1, s1, s5
+; GFX11-NEXT: s_lshr_b32 s5, s3, 16
+; GFX11-NEXT: s_lshl_b32 s2, s2, 0x80008
+; GFX11-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-NEXT: s_lshl_b32 s3, s3, 0x80008
+; GFX11-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-NEXT: s_pack_ll_b32_b16 s2, s2, s4
+; GFX11-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-NEXT: s_pack_ll_b32_b16 s3, s3, s5
+; GFX11-NEXT: s_lshr_b32 s5, s1, 16
+; GFX11-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-NEXT: v_pk_add_u16 v0, s2, s3 clamp
+; GFX11-NEXT: s_pack_ll_b32_b16 s0, s0, s4
+; GFX11-NEXT: s_pack_ll_b32_b16 s1, s1, s5
+; GFX11-NEXT: v_pk_add_u16 v1, s0, s1 clamp
+; GFX11-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-NEXT: v_readfirstlane_b32 s1, v1
+; GFX11-NEXT: s_and_b32 s2, s0, 0xffff
+; GFX11-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-NEXT: s_and_b32 s3, s1, 0xffff
+; GFX11-NEXT: s_lshr_b32 s1, s1, 16
+; GFX11-NEXT: s_lshr_b32 s3, s3, 0x80008
+; GFX11-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-NEXT: s_pack_ll_b32_b16 s0, s2, s0
+; GFX11-NEXT: s_pack_ll_b32_b16 s1, s3, s1
+; GFX11-NEXT: s_bfe_u32 s2, s0, 0x80010
+; GFX11-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-NEXT: s_and_b32 s3, s1, 0xff
+; GFX11-NEXT: s_or_b32 s0, s0, s2
+; GFX11-NEXT: s_lshl_b32 s2, s3, 16
+; GFX11-NEXT: s_bfe_u32 s1, s1, 0x80010
+; GFX11-NEXT: s_or_b32 s0, s0, s2
+; GFX11-NEXT: s_lshl_b32 s1, s1, 24
+; GFX11-NEXT: s_or_b32 s0, s0, s1
+; GFX11-NEXT: ; return to shader part epilog
%lhs = bitcast i32 %lhs.arg to <4 x i8>
%rhs = bitcast i32 %rhs.arg to <4 x i8>
%result = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> %lhs, <4 x i8> %rhs)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll
index 1c8a26ce6126d..8f248b4d2700c 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll
@@ -334,8 +334,11 @@ define i16 @v_usubsat_v2i8(i16 %lhs.arg, i16 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, v0, v1 clamp
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v0 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
-; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_usubsat_v2i8:
@@ -434,86 +437,33 @@ define amdgpu_ps i16 @s_usubsat_v2i8(i16 inreg %lhs.arg, i16 inreg %rhs.arg) {
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
-; GFX10-LABEL: s_usubsat_v2i8:
-; GFX10: ; %bb.0:
-; GFX10-NEXT: s_lshr_b32 s2, s0, 8
-; GFX10-NEXT: s_lshr_b32 s3, s1, 8
-; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10-NEXT: s_lshr_b32 s2, s0, 16
-; GFX10-NEXT: s_lshr_b32 s3, s1, 16
-; GFX10-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX10-NEXT: s_lshl_b32 s2, s2, 8
-; GFX10-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX10-NEXT: s_lshl_b32 s3, s3, 8
-; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
-; GFX10-NEXT: v_readfirstlane_b32 s0, v0
-; GFX10-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX10-NEXT: s_lshr_b32 s0, s0, 16
-; GFX10-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX10-NEXT: s_lshr_b32 s0, s0, 8
-; GFX10-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX10-NEXT: s_lshr_b32 s1, s0, 16
-; GFX10-NEXT: s_and_b32 s0, s0, 0xff
-; GFX10-NEXT: s_and_b32 s1, s1, 0xff
-; GFX10-NEXT: s_lshl_b32 s1, s1, 8
-; GFX10-NEXT: s_or_b32 s0, s0, s1
-; GFX10-NEXT: ; return to shader part epilog
-;
-; GFX11-TRUE16-LABEL: s_usubsat_v2i8:
-; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-TRUE16-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s0, 8
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-TRUE16-NEXT: ; return to shader part epilog
-;
-; GFX11-FAKE16-LABEL: s_usubsat_v2i8:
-; GFX11-FAKE16: ; %bb.0:
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX11-FAKE16-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-FAKE16-NEXT: ; return to shader part epilog
+; GFX10PLUS-LABEL: s_usubsat_v2i8:
+; GFX10PLUS: ; %bb.0:
+; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 8
+; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 8
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 16
+; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 16
+; GFX10PLUS-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX10PLUS-NEXT: s_lshl_b32 s2, s2, 8
+; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX10PLUS-NEXT: s_lshl_b32 s3, s3, 8
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10PLUS-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
+; GFX10PLUS-NEXT: v_readfirstlane_b32 s0, v0
+; GFX10PLUS-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 16
+; GFX10PLUS-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 8
+; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX10PLUS-NEXT: s_lshr_b32 s1, s0, 16
+; GFX10PLUS-NEXT: s_and_b32 s0, s0, 0xff
+; GFX10PLUS-NEXT: s_and_b32 s1, s1, 0xff
+; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 8
+; GFX10PLUS-NEXT: s_or_b32 s0, s0, s1
+; GFX10PLUS-NEXT: ; return to shader part epilog
%lhs = bitcast i16 %lhs.arg to <2 x i8>
%rhs = bitcast i16 %rhs.arg to <2 x i8>
%result = call <2 x i8> @llvm.usub.sat.v2i8(<2 x i8> %lhs, <2 x i8> %rhs)
@@ -667,21 +617,19 @@ define i32 @v_usubsat_v4i8(i32 %lhs.arg, i32 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v0, 8, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v2, 8, v3 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v3, 8, v6 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, v0, v1 clamp
-; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v6 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_sub_u16 v1, v2, v3 clamp
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v0, 8, v0 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_sub_u16 v1, v2, v1 clamp
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xff, v2
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v1.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 8
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v1, v1, 16, 8
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v3
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 24, v1
; GFX11-TRUE16-NEXT: v_and_or_b32 v0, 0xff, v0, v2
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 24, v3
-; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v2
+; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v3, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_usubsat_v4i8:
@@ -912,107 +860,57 @@ define amdgpu_ps i32 @s_usubsat_v4i8(i32 inreg %lhs.arg, i32 inreg %rhs.arg) {
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
-; GFX11-TRUE16-LABEL: s_usubsat_v4i8:
-; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s0, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 24
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
-; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s2, 16
-; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s3, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, s2, s3 clamp
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
-; GFX11-TRUE16-NEXT: v_pk_sub_u16 v1, s0, s1 clamp
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s1, v1
-; GFX11-TRUE16-NEXT: s_and_b32 s2, s0, 0xffff
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
-; GFX11-TRUE16-NEXT: s_and_b32 s2, s1, 0xffff
-; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
-; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s2, s1
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s0, 8
-; GFX11-TRUE16-NEXT: s_and_b32 s1, s1, 0xff
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s1, 16
-; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 24
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-TRUE16-NEXT: ; return to shader part epilog
-;
-; GFX11-FAKE16-LABEL: s_usubsat_v4i8:
-; GFX11-FAKE16: ; %bb.0:
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s0, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 24
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
-; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s2, 16
-; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s3, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-FAKE16-NEXT: v_pk_sub_u16 v0, s2, s3 clamp
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
-; GFX11-FAKE16-NEXT: v_pk_sub_u16 v1, s0, s1 clamp
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s1, v1
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s0, 0xffff
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xffff
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s3, 0x80008
-; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 8
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s3, s1
-; GFX11-FAKE16-NEXT: s_bfe_u32 s2, s0, 0x80010
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s3, 16
-; GFX11-FAKE16-NEXT: s_bfe_u32 s1, s1, 0x80010
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 24
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-FAKE16-NEXT: ; return to shader part epilog
+; GFX11-LABEL: s_usubsat_v4i8:
+; GFX11: ; %bb.0:
+; GFX11-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-NEXT: s_lshr_b32 s4, s1, 8
+; GFX11-NEXT: s_lshr_b32 s5, s1, 24
+; GFX11-NEXT: s_pack_ll_b32_b16 s2, s0, s2
+; GFX11-NEXT: s_pack_hl_b32_b16 s0, s0, s3
+; GFX11-NEXT: s_pack_ll_b32_b16 s3, s1, s4
+; GFX11-NEXT: s_lshr_b32 s4, s2, 16
+; GFX11-NEXT: s_pack_hl_b32_b16 s1, s1, s5
+; GFX11-NEXT: s_lshr_b32 s5, s3, 16
+; GFX11-NEXT: s_lshl_b32 s2, s2, 0x80008
+; GFX11-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-NEXT: s_lshl_b32 s3, s3, 0x80008
+; GFX11-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-NEXT: s_pack_ll_b32_b16 s2, s2, s4
+; GFX11-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-NEXT: s_pack_ll_b32_b16 s3, s3, s5
+; GFX11-NEXT: s_lshr_b32 s5, s1, 16
+; GFX11-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-NEXT: v_pk_sub_u16 v0, s2, s3 clamp
+; GFX11-NEXT: s_pack_ll_b32_b16 s0, s0, s4
+; GFX11-NEXT: s_pack_ll_b32_b16 s1, s1, s5
+; GFX11-NEXT: v_pk_sub_u16 v1, s0, s1 clamp
+; GFX11-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-NEXT: v_readfirstlane_b32 s1, v1
+; GFX11-NEXT: s_and_b32 s2, s0, 0xffff
+; GFX11-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-NEXT: s_and_b32 s3, s1, 0xffff
+; GFX11-NEXT: s_lshr_b32 s1, s1, 16
+; GFX11-NEXT: s_lshr_b32 s3, s3, 0x80008
+; GFX11-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-NEXT: s_pack_ll_b32_b16 s0, s2, s0
+; GFX11-NEXT: s_pack_ll_b32_b16 s1, s3, s1
+; GFX11-NEXT: s_bfe_u32 s2, s0, 0x80010
+; GFX11-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-NEXT: s_and_b32 s3, s1, 0xff
+; GFX11-NEXT: s_or_b32 s0, s0, s2
+; GFX11-NEXT: s_lshl_b32 s2, s3, 16
+; GFX11-NEXT: s_bfe_u32 s1, s1, 0x80010
+; GFX11-NEXT: s_or_b32 s0, s0, s2
+; GFX11-NEXT: s_lshl_b32 s1, s1, 24
+; GFX11-NEXT: s_or_b32 s0, s0, s1
+; GFX11-NEXT: ; return to shader part epilog
%lhs = bitcast i32 %lhs.arg to <4 x i8>
%rhs = bitcast i32 %rhs.arg to <4 x i8>
%result = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> %lhs, <4 x i8> %rhs)
diff --git a/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll b/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll
index 9db34d0f1ed1b..9c5850a03c9ab 100644
--- a/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll
+++ b/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll
@@ -410,14 +410,25 @@ define void @undef_lo2_v4i16(<2 x i16> %arg0) {
; GFX11-FAKE16-NEXT: ;;#ASMEND
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-TRUE16-LABEL: undef_lo2_v4i16:
-; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: ;;#ASMSTART
-; GFX11-TRUE16-NEXT: ; use v[0:1]
-; GFX11-TRUE16-NEXT: ;;#ASMEND
-; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-SDAG-LABEL: undef_lo2_v4i16:
+; GFX11-TRUE16-SDAG: ; %bb.0:
+; GFX11-TRUE16-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-SDAG-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX11-TRUE16-SDAG-NEXT: ;;#ASMSTART
+; GFX11-TRUE16-SDAG-NEXT: ; use v[0:1]
+; GFX11-TRUE16-SDAG-NEXT: ;;#ASMEND
+; GFX11-TRUE16-SDAG-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-TRUE16-GISEL-LABEL: undef_lo2_v4i16:
+; GFX11-TRUE16-GISEL: ; %bb.0:
+; GFX11-TRUE16-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-GISEL-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-TRUE16-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-GISEL-NEXT: v_mov_b16_e32 v0.h, v0.l
+; GFX11-TRUE16-GISEL-NEXT: ;;#ASMSTART
+; GFX11-TRUE16-GISEL-NEXT: ; use v[0:1]
+; GFX11-TRUE16-GISEL-NEXT: ;;#ASMEND
+; GFX11-TRUE16-GISEL-NEXT: s_setpc_b64 s[30:31]
%undef.lo = shufflevector <2 x i16> %arg0, <2 x i16> poison, <4 x i32> <i32 1, i32 1, i32 2, i32 3>
call void asm sideeffect "; use $0", "v"(<4 x i16> %undef.lo);
ret void
@@ -464,14 +475,25 @@ define void @undef_lo2_v4f16(<2 x half> %arg0) {
; GFX11-FAKE16-NEXT: ;;#ASMEND
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-TRUE16-LABEL: undef_lo2_v4f16:
-; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: ;;#ASMSTART
-; GFX11-TRUE16-NEXT: ; use v[0:1]
-; GFX11-TRUE16-NEXT: ;;#ASMEND
-; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-SDAG-LABEL: undef_lo2_v4f16:
+; GFX11-TRUE16-SDAG: ; %bb.0:
+; GFX11-TRUE16-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-SDAG-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX11-TRUE16-SDAG-NEXT: ;;#ASMSTART
+; GFX11-TRUE16-SDAG-NEXT: ; use v[0:1]
+; GFX11-TRUE16-SDAG-NEXT: ;;#ASMEND
+; GFX11-TRUE16-SDAG-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX11-TRUE16-GISEL-LABEL: undef_lo2_v4f16:
+; GFX11-TRUE16-GISEL: ; %bb.0:
+; GFX11-TRUE16-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-GISEL-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-TRUE16-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-GISEL-NEXT: v_mov_b16_e32 v0.h, v0.l
+; GFX11-TRUE16-GISEL-NEXT: ;;#ASMSTART
+; GFX11-TRUE16-GISEL-NEXT: ; use v[0:1]
+; GFX11-TRUE16-GISEL-NEXT: ;;#ASMEND
+; GFX11-TRUE16-GISEL-NEXT: s_setpc_b64 s[30:31]
%undef.lo = shufflevector <2 x half> %arg0, <2 x half> poison, <4 x i32> <i32 1, i32 1, i32 2, i32 3>
call void asm sideeffect "; use $0", "v"(<4 x half> %undef.lo);
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll b/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll
index 76583e806b805..9105358e7bdbf 100644
--- a/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll
+++ b/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll
@@ -1865,13 +1865,14 @@ define <2 x half> @fmul_select_v2f16_test3(<2 x half> %x, <2 x i32> %bool.arg1,
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v4
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: fmul_select_v2f16_test3:
@@ -2031,13 +2032,14 @@ define <2 x half> @fmul_select_v2f16_test4(<2 x half> %x, <2 x i32> %bool.arg1,
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, -1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v4
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, -1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: fmul_select_v2f16_test4:
diff --git a/llvm/test/CodeGen/AMDGPU/fma.f16.ll b/llvm/test/CodeGen/AMDGPU/fma.f16.ll
index 8d1d08d1e7c68..df9e86756815f 100644
--- a/llvm/test/CodeGen/AMDGPU/fma.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fma.f16.ll
@@ -653,12 +653,14 @@ define <2 x i32> @test_D139469_v2f16(<2 x half> %arg) {
; GFX11-GISEL-TRUE16-NEXT: v_pk_mul_f16 v2, 0x291e291e, v0
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_pk_fma_f16 v0, 0x291e291e, v0, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e32 vcc_lo, 0, v2.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v2.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s0, 0, v0.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v1.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v3.l
; GFX11-GISEL-TRUE16-NEXT: s_or_b32 s0, vcc_lo, s0
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
; GFX11-GISEL-TRUE16-NEXT: s_or_b32 s0, s1, s2
@@ -744,12 +746,14 @@ define <2 x i32> @test_D139469_v2f16(<2 x half> %arg) {
; GFX12-GISEL-TRUE16-NEXT: v_pk_mul_f16 v2, 0x291e291e, v0
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-GISEL-TRUE16-NEXT: v_pk_fma_f16 v0, 0x291e291e, v0, v1
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e32 vcc_lo, 0, v2.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v2.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s0, 0, v0.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v1.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v3.l
; GFX12-GISEL-TRUE16-NEXT: s_or_b32 s0, vcc_lo, s0
; GFX12-GISEL-TRUE16-NEXT: s_wait_alu depctr_sa_sdst(0)
; GFX12-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
diff --git a/llvm/test/CodeGen/AMDGPU/fmed3.ll b/llvm/test/CodeGen/AMDGPU/fmed3.ll
index 668347eb97004..e5b533ec1d60d 100644
--- a/llvm/test/CodeGen/AMDGPU/fmed3.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmed3.ll
@@ -9194,19 +9194,22 @@ define <2 x half> @v_test_nnan_input_fmed3_r_i_i_v2f16_maximum_minimum(<2 x half
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, 1.0 op_sel_hi:[1,0]
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_pk_max_f16 v1, v0, 2.0 op_sel_hi:[1,0]
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_pk_max_f16 v2, v0, 2.0 op_sel_hi:[1,0]
; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e32 vcc_lo, 2.0, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 2.0, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v1.l, vcc_lo
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 2.0, v1.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v1.l, 0x7e00, v2.l, vcc_lo
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v1.h, 0x7e00, v0.l, s0
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.h, 0x7e00, v1.h, s0
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e32 vcc_lo, 4.0, v0.l
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e32 vcc_lo, 4.0, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_pk_min_f16 v0, v1, 4.0 op_sel_hi:[1,0]
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 4.0, v1.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_pk_min_f16 v1, v0, 4.0 op_sel_hi:[1,0]
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 4.0, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v1.l, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.h, 0x7e00, v1.h, s0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v0.l, vcc_lo
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.h, 0x7e00, v1.l, s0
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
%a.add = fadd nnan <2 x half> %a, splat (half 1.0)
%max = call <2 x half> @llvm.maximum.v2f16(<2 x half> %a.add, <2 x half> splat (half 2.0))
diff --git a/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll b/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll
index 95d2f07402dd4..8613d238d4201 100644
--- a/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll
@@ -6793,11 +6793,12 @@ define <2 x half> @v_mul_v2f16_select_64_1(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 6, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 6, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_64_1:
@@ -6897,11 +6898,12 @@ define <2 x half> @v_mul_v2f16_select_1_64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_1_64:
@@ -7006,9 +7008,10 @@ define <2 x half> @v_mul_v2f16_select_n1_n64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_n1_n64:
@@ -7120,6 +7123,7 @@ define <2 x half> @v_mul_v2f16_select_128_64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_dual_mov_b32 v3, 0x7fff :: v_dual_add_nc_u32 v0, 6, v0
@@ -7128,7 +7132,7 @@ define <2 x half> @v_mul_v2f16_select_128_64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v1, 6, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_128_64:
@@ -7249,13 +7253,15 @@ define <2 x half> @v_mul_v2f16_select_n128_n64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-GISEL-TRUE16-NEXT: v_dual_mov_b32 v3, 0x7fff :: v_dual_add_nc_u32 v0, 6, v0
+; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v0, 6, v0
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v0, 0xffff8000, v0, v3
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v1, 6, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_n128_n64:
@@ -7366,9 +7372,10 @@ define <2 x half> @v_mul_v2f16_select_n128_n16(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 4, 7, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 4, 7, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_n128_n16:
@@ -7471,11 +7478,12 @@ define <2 x half> @v_contract_mul_add_v2f16_select_64_1(<2 x i32> %arg, <2 x hal
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 6, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 6, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7580,11 +7588,12 @@ define <2 x half> @v_contract_mul_add_v2f16_select_1_64(<2 x i32> %arg, <2 x hal
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7694,9 +7703,10 @@ define <2 x half> @v_contract_mul_add_v2f16_select_n64_n1(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 6, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 6, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7807,9 +7817,10 @@ define <2 x half> @v_contract_mul_add_v2f16_select_n1_n64(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7926,6 +7937,7 @@ define <2 x half> @v_contract_mul_add_v2f16_select_128_64(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v0, 6, v0
@@ -7934,7 +7946,7 @@ define <2 x half> @v_contract_mul_add_v2f16_select_128_64(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v0, 0xffff8000, v0, v4
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v4
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v5.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -8043,11 +8055,12 @@ define <2 x half> @v_contract_mul_add_v2f16_select_128_4(<2 x i32> %arg, <2 x ha
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 2, 7, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 2, 7, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -8163,6 +8176,7 @@ define <2 x half> @v_contract_mul_add_v2f16_select_2_4(<2 x i32> %arg, <2 x half
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v0, 2, v0
@@ -8171,7 +8185,7 @@ define <2 x half> @v_contract_mul_add_v2f16_select_2_4(<2 x i32> %arg, <2 x half
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v0, 0xffff8000, v0, v4
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v4
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v5.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -8280,11 +8294,12 @@ define <2 x half> @v_contract_mul_add_v2f16_select_4_128(<2 x i32> %arg, <2 x ha
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 7, 2, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 7, 2, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
index eb1d950a90467..7450f1c6d4b37 100644
--- a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
@@ -4205,12 +4205,28 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16lo_sexti8_reg_hi(ptr addrs
; GFX11-NEXT: v_mov_b32_e32 v0, v1
; GFX11-NEXT: ; return to shader part epilog
;
-; GFX12-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi:
-; GFX12: ; %bb.0:
-; GFX12-NEXT: global_load_d16_i8 v1, v0, s[2:3]
-; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_mov_b32_e32 v0, v1
-; GFX12-NEXT: ; return to shader part epilog
+; GFX12-SDAG-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi:
+; GFX12-SDAG: ; %bb.0:
+; GFX12-SDAG-NEXT: global_load_d16_i8 v1, v0, s[2:3]
+; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
+; GFX12-SDAG-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-SDAG-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi:
+; GFX12-GISEL-TRUE16: ; %bb.0:
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_i8 v0, v0, s[2:3]
+; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi:
+; GFX12-GISEL-FAKE16: ; %bb.0:
+; GFX12-GISEL-FAKE16-NEXT: global_load_d16_i8 v1, v0, s[2:3]
+; GFX12-GISEL-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-FAKE16-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
%zext.offset = zext i32 %voffset to i64
%gep0 = getelementptr inbounds i8, ptr addrspace(1) %sbase, i64 %zext.offset
%load = load i8, ptr addrspace(1) %gep0
@@ -4235,12 +4251,28 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128
; GFX11-NEXT: v_mov_b32_e32 v0, v1
; GFX11-NEXT: ; return to shader part epilog
;
-; GFX12-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
-; GFX12: ; %bb.0:
-; GFX12-NEXT: global_load_d16_i8 v1, v0, s[2:3] offset:-128
-; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_mov_b32_e32 v0, v1
-; GFX12-NEXT: ; return to shader part epilog
+; GFX12-SDAG-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
+; GFX12-SDAG: ; %bb.0:
+; GFX12-SDAG-NEXT: global_load_d16_i8 v1, v0, s[2:3] offset:-128
+; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
+; GFX12-SDAG-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-SDAG-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
+; GFX12-GISEL-TRUE16: ; %bb.0:
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_i8 v0, v0, s[2:3] offset:-128
+; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
+; GFX12-GISEL-FAKE16: ; %bb.0:
+; GFX12-GISEL-FAKE16-NEXT: global_load_d16_i8 v1, v0, s[2:3] offset:-128
+; GFX12-GISEL-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-FAKE16-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
%zext.offset = zext i32 %voffset to i64
%gep0 = getelementptr inbounds i8, ptr addrspace(1) %sbase, i64 %zext.offset
%gep1 = getelementptr inbounds i8, ptr addrspace(1) %gep0, i64 -128
@@ -4594,12 +4626,29 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16hi_sexti8_reg_hi(ptr addrs
; GFX11-NEXT: v_mov_b32_e32 v0, v1
; GFX11-NEXT: ; return to shader part epilog
;
-; GFX12-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi:
-; GFX12: ; %bb.0:
-; GFX12-NEXT: global_load_d16_hi_i8 v1, v0, s[2:3]
-; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_mov_b32_e32 v0, v1
-; GFX12-NEXT: ; return to shader part epilog
+; GFX12-SDAG-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi:
+; GFX12-SDAG: ; %bb.0:
+; GFX12-SDAG-NEXT: global_load_d16_hi_i8 v1, v0, s[2:3]
+; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
+; GFX12-SDAG-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-SDAG-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi:
+; GFX12-GISEL-TRUE16: ; %bb.0:
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_i8 v0, v0, s[2:3]
+; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi:
+; GFX12-GISEL-FAKE16: ; %bb.0:
+; GFX12-GISEL-FAKE16-NEXT: global_load_d16_hi_i8 v1, v0, s[2:3]
+; GFX12-GISEL-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-FAKE16-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
%zext.offset = zext i32 %voffset to i64
%gep0 = getelementptr inbounds i8, ptr addrspace(1) %sbase, i64 %zext.offset
%load = load i8, ptr addrspace(1) %gep0
@@ -4624,12 +4673,29 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128
; GFX11-NEXT: v_mov_b32_e32 v0, v1
; GFX11-NEXT: ; return to shader part epilog
;
-; GFX12-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
-; GFX12: ; %bb.0:
-; GFX12-NEXT: global_load_d16_hi_i8 v1, v0, s[2:3] offset:-128
-; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_mov_b32_e32 v0, v1
-; GFX12-NEXT: ; return to shader part epilog
+; GFX12-SDAG-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
+; GFX12-SDAG: ; %bb.0:
+; GFX12-SDAG-NEXT: global_load_d16_hi_i8 v1, v0, s[2:3] offset:-128
+; GFX12-SDAG-NEXT: s_wait_loadcnt 0x0
+; GFX12-SDAG-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-SDAG-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
+; GFX12-GISEL-TRUE16: ; %bb.0:
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_i8 v0, v0, s[2:3] offset:-128
+; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
+; GFX12-GISEL-FAKE16: ; %bb.0:
+; GFX12-GISEL-FAKE16-NEXT: global_load_d16_hi_i8 v1, v0, s[2:3] offset:-128
+; GFX12-GISEL-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX12-GISEL-FAKE16-NEXT: v_mov_b32_e32 v0, v1
+; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
%zext.offset = zext i32 %voffset to i64
%gep0 = getelementptr inbounds i8, ptr addrspace(1) %sbase, i64 %zext.offset
%gep1 = getelementptr inbounds i8, ptr addrspace(1) %gep0, i64 -128
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
index 614566a230f68..a89de01b2354c 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
@@ -114,45 +114,25 @@ define amdgpu_ps <4 x float> @image_bvh_intersect_ray_a16(i32 inreg %node_ptr, f
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
; GFX12-SDAG-NEXT: ; return to shader part epilog
;
-; GFX12-GISEL-TRUE16-LABEL: image_bvh_intersect_ray_a16:
-; GFX12-GISEL-TRUE16: ; %bb.0: ; %main_body
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s20, s2
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s22, s4
-; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s4, s7, s5
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s21, s3
-; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s8, s6
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s5, s4
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v1, s21
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v3, s4
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v6, s0 :: v_dual_mov_b32 v7, s1
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s6
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s16, s9
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s17, s10
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s18, s11
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s19, s12
-; GFX12-GISEL-TRUE16-NEXT: image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[16:19] a16
-; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
-;
-; GFX12-GISEL-FAKE16-LABEL: image_bvh_intersect_ray_a16:
-; GFX12-GISEL-FAKE16: ; %bb.0: ; %main_body
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s20, s2
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s22, s4
-; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s7, s5
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s21, s3
-; GFX12-GISEL-FAKE16-NEXT: s_pack_hh_b32_b16 s5, s7, s5
-; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s6, s8, s6
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v1, s21
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v3, s4
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v6, s0 :: v_dual_mov_b32 v7, s1
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s6
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s16, s9
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s17, s10
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s18, s11
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s19, s12
-; GFX12-GISEL-FAKE16-NEXT: image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[16:19] a16
-; GFX12-GISEL-FAKE16-NEXT: s_wait_bvhcnt 0x0
-; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
+; GFX12-GISEL-LABEL: image_bvh_intersect_ray_a16:
+; GFX12-GISEL: ; %bb.0: ; %main_body
+; GFX12-GISEL-NEXT: s_mov_b32 s20, s2
+; GFX12-GISEL-NEXT: s_mov_b32 s22, s4
+; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s4, s7, s5
+; GFX12-GISEL-NEXT: s_mov_b32 s21, s3
+; GFX12-GISEL-NEXT: s_pack_hh_b32_b16 s5, s7, s5
+; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s6, s8, s6
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v1, s21
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v3, s4
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v6, s0 :: v_dual_mov_b32 v7, s1
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s6
+; GFX12-GISEL-NEXT: s_mov_b32 s16, s9
+; GFX12-GISEL-NEXT: s_mov_b32 s17, s10
+; GFX12-GISEL-NEXT: s_mov_b32 s18, s11
+; GFX12-GISEL-NEXT: s_mov_b32 s19, s12
+; GFX12-GISEL-NEXT: image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[16:19] a16
+; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
+; GFX12-GISEL-NEXT: ; return to shader part epilog
main_body:
%v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i32.v4f16(i32 %node_ptr, float %ray_extent, <3 x float> %ray_origin, <3 x half> %ray_dir, <3 x half> %ray_inv_dir, <4 x i32> %tdescr)
%r = bitcast <4 x i32> %v to <4 x float>
@@ -256,47 +236,26 @@ define amdgpu_ps <4 x float> @image_bvh64_intersect_ray_a16(i64 inreg %node_ptr,
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
; GFX12-SDAG-NEXT: ; return to shader part epilog
;
-; GFX12-GISEL-TRUE16-LABEL: image_bvh64_intersect_ray_a16:
-; GFX12-GISEL-TRUE16: ; %bb.0: ; %main_body
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s21, s4
-; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s4, s8, s6
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s20, s3
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s22, s5
-; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s9, s7
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s5, s4
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v3, s4
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v1, s21
-; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v5, s6
-; GFX12-GISEL-TRUE16-NEXT: v_mov_b32_e32 v4, s5
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s16, s10
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s17, s11
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s18, s12
-; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s19, s13
-; GFX12-GISEL-TRUE16-NEXT: image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[16:19] a16
-; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
-;
-; GFX12-GISEL-FAKE16-LABEL: image_bvh64_intersect_ray_a16:
-; GFX12-GISEL-FAKE16: ; %bb.0: ; %main_body
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s20, s3
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s21, s4
-; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s8, s6
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s22, s5
-; GFX12-GISEL-FAKE16-NEXT: s_pack_hh_b32_b16 s5, s8, s6
-; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s6, s9, s7
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v3, s4
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v1, s21
-; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v5, s6
-; GFX12-GISEL-FAKE16-NEXT: v_mov_b32_e32 v4, s5
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s16, s10
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s17, s11
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s18, s12
-; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s19, s13
-; GFX12-GISEL-FAKE16-NEXT: image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[16:19] a16
-; GFX12-GISEL-FAKE16-NEXT: s_wait_bvhcnt 0x0
-; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
+; GFX12-GISEL-LABEL: image_bvh64_intersect_ray_a16:
+; GFX12-GISEL: ; %bb.0: ; %main_body
+; GFX12-GISEL-NEXT: s_mov_b32 s20, s3
+; GFX12-GISEL-NEXT: s_mov_b32 s21, s4
+; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s4, s8, s6
+; GFX12-GISEL-NEXT: s_mov_b32 s22, s5
+; GFX12-GISEL-NEXT: s_pack_hh_b32_b16 s5, s8, s6
+; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s6, s9, s7
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v3, s4
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v1, s21
+; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v5, s6
+; GFX12-GISEL-NEXT: v_mov_b32_e32 v4, s5
+; GFX12-GISEL-NEXT: s_mov_b32 s16, s10
+; GFX12-GISEL-NEXT: s_mov_b32 s17, s11
+; GFX12-GISEL-NEXT: s_mov_b32 s18, s12
+; GFX12-GISEL-NEXT: s_mov_b32 s19, s13
+; GFX12-GISEL-NEXT: image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[16:19] a16
+; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
+; GFX12-GISEL-NEXT: ; return to shader part epilog
main_body:
%v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i64.v4f16(i64 %node_ptr, float %ray_extent, <3 x float> %ray_origin, <3 x half> %ray_dir, <3 x half> %ray_inv_dir, <4 x i32> %tdescr)
%r = bitcast <4 x i32> %v to <4 x float>
@@ -932,5 +891,7 @@ declare i32 @llvm.amdgcn.workitem.id.x()
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX11-FAKE16: {{.*}}
; GFX11-TRUE16: {{.*}}
+; GFX12-GISEL-FAKE16: {{.*}}
+; GFX12-GISEL-TRUE16: {{.*}}
; GFX12-SDAG-FAKE16: {{.*}}
; GFX12-SDAG-TRUE16: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll b/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll
index 7300e4227925f..6510824ec3ea6 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll
@@ -535,13 +535,15 @@ define { <2 x half>, <2 x i32> } @test_frexp_v2f16_v2i32(<2 x half> %a) {
; GFX11-GISEL-TRUE16-LABEL: test_frexp_v2f16_v2i32:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
-; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v2, 0, 16
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v3.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v2, 0, 16
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v3, 0, 16
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32:
@@ -566,13 +568,15 @@ define { <2 x half>, <2 x i32> } @test_frexp_v2f16_v2i32(<2 x half> %a) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.l
; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
-; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v2, 0, 16
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v3.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v2, 0, 16
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v3, 0, 16
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32:
@@ -705,8 +709,10 @@ define <2 x half> @test_frexp_v2f16_v2i32_only_use_fract(<2 x half> %a) {
; GFX11-GISEL-TRUE16-LABEL: test_frexp_v2f16_v2i32_only_use_fract:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32_only_use_fract:
@@ -726,8 +732,10 @@ define <2 x half> @test_frexp_v2f16_v2i32_only_use_fract(<2 x half> %a) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32_only_use_fract:
@@ -872,10 +880,12 @@ define <2 x i32> @test_frexp_v2f16_v2i32_only_use_exp(<2 x half> %a) {
; GFX11-GISEL-TRUE16-LABEL: test_frexp_v2f16_v2i32_only_use_exp:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v0, v0, 0, 16
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -898,10 +908,12 @@ define <2 x i32> @test_frexp_v2f16_v2i32_only_use_exp(<2 x half> %a) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v0, v0, 0, 16
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
index cc5f6f842625c..3b5763b784184 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
@@ -1446,11 +1446,10 @@ define <2 x i1> @isnan_v2f16(<2 x half> %x) nounwind {
; GFX11GLISEL-TRUE16: ; %bb.0:
; GFX11GLISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v1, 3 :: v_dual_mov_b32 v2, 3
+; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v0
; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.l, v1.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
-; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.h, v3.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v0, v2
+; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v3.l, v2.l
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX11GLISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -1609,14 +1608,13 @@ define <3 x i1> @isnan_v3f16(<3 x half> %x) nounwind {
; GFX11GLISEL-TRUE16: ; %bb.0:
; GFX11GLISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v2, 3 :: v_dual_mov_b32 v3, 3
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v4, 3
+; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v0
; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.l, v2.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.l
-; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v4, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.h, v3.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v0, v4
+; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v5.l, v3.l
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v5.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v4, 3
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v4.l
; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v1, v3
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11GLISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -1805,18 +1803,19 @@ define <4 x i1> @isnan_v4f16(<4 x half> %x) nounwind {
; GFX11GLISEL-TRUE16: ; %bb.0:
; GFX11GLISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v2, 3 :: v_dual_mov_b32 v3, 3
-; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v4, 3 :: v_dual_mov_b32 v5, 3
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v4, 3
+; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v0
+; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v1
; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.l, v2.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v6.l, v4.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v7.l, v5.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.l
+; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v7.l, v3.l
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v4, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.h, v3.l
-; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v5, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v6.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v0, v4
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v6, 3
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v5.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v1, v4
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.h, v7.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v8.l, v6.l
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo
; GFX11GLISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll b/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll
index 43cc632d3708b..dcda6583527e4 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll
@@ -512,12 +512,13 @@ define <2 x half> @test_ldexp_v2f16_v2i32(<2 x half> %a, <2 x i32> %b) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v3
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v2f16_v2i32:
@@ -625,8 +626,11 @@ define <2 x half> @test_ldexp_v2f16_v2i16(<2 x half> %a, <2 x i16> %b) {
; GFX11-GISEL-TRUE16-LABEL: test_ldexp_v2f16_v2i16:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v2f16_v2i16:
@@ -771,13 +775,14 @@ define <3 x half> @test_ldexp_v3f16_v3i32(<3 x half> %a, <3 x i32> %b) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v5, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v5
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v3, v5
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v5
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v6.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v4.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -906,9 +911,12 @@ define <3 x half> @test_ldexp_v3f16_v3i16(<3 x half> %a, <3 x i16> %b) {
; GFX11-GISEL-TRUE16-LABEL: test_ldexp_v3f16_v3i16:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v3f16_v3i16:
@@ -1088,17 +1096,19 @@ define <4 x half> @test_ldexp_v4f16_v4i32(<4 x half> %a, <4 x i32> %b) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v6, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v1
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v6
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v3, v6
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v6
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v5, v6
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v7.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v4.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v1.h, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v8.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v4f16_v4i32:
@@ -1257,10 +1267,15 @@ define <4 x half> @test_ldexp_v4f16_v4i16(<4 x half> %a, <4 x i16> %b) {
; GFX11-GISEL-TRUE16-LABEL: test_ldexp_v4f16_v4i16:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v1.h, v3.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v6.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v5.l, v7.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v4f16_v4i16:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log.ll b/llvm/test/CodeGen/AMDGPU/llvm.log.ll
index 7903ae93d770c..0c5adcbb8c4df 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log.ll
@@ -6722,8 +6722,10 @@ define <2 x half> @v_log_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v2f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -6860,9 +6862,10 @@ define <2 x half> @v_log_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7003,9 +7006,10 @@ define <2 x half> @v_log_fneg_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_or_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7147,9 +7151,10 @@ define <2 x half> @v_log_fneg_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7275,8 +7280,10 @@ define <2 x half> @v_log_v2f16_fast(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v2f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7386,14 +7393,15 @@ define <3 x half> @v_log_v3f16(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v3f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v3f16:
@@ -7503,14 +7511,15 @@ define <3 x half> @v_log_v3f16_fast(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v3f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v3f16_fast:
@@ -7698,16 +7707,19 @@ define <4 x half> @v_log_v4f16(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v4f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v4f16:
@@ -7900,16 +7912,19 @@ define <4 x half> @v_log_v4f16_fast(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v4f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v4f16_fast:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log10.ll b/llvm/test/CodeGen/AMDGPU/llvm.log10.ll
index 478580ff8ec0a..e2fcb18ee3217 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log10.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log10.ll
@@ -6722,8 +6722,10 @@ define <2 x half> @v_log10_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v2f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -6860,9 +6862,10 @@ define <2 x half> @v_log10_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7003,9 +7006,10 @@ define <2 x half> @v_log10_fneg_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_or_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7147,9 +7151,10 @@ define <2 x half> @v_log10_fneg_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7275,8 +7280,10 @@ define <2 x half> @v_log10_v2f16_fast(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v2f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7386,14 +7393,15 @@ define <3 x half> @v_log10_v3f16(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v3f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v3f16:
@@ -7503,14 +7511,15 @@ define <3 x half> @v_log10_v3f16_fast(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v3f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v3f16_fast:
@@ -7698,16 +7707,19 @@ define <4 x half> @v_log10_v4f16(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v4f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v4f16:
@@ -7900,16 +7912,19 @@ define <4 x half> @v_log10_v4f16_fast(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v4f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v4f16_fast:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log2.ll b/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
index 8401e05b39c19..0abf90907d2f9 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
@@ -4094,8 +4094,10 @@ define <2 x half> @v_log2_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v2f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v2f16:
@@ -4208,9 +4210,10 @@ define <2 x half> @v_log2_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_fabs_v2f16:
@@ -4311,9 +4314,10 @@ define <2 x half> @v_log2_fneg_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_or_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_fneg_fabs_v2f16:
@@ -4415,9 +4419,10 @@ define <2 x half> @v_log2_fneg_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_fneg_v2f16:
@@ -4514,8 +4519,10 @@ define <2 x half> @v_log2_v2f16_fast(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v2f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v2f16_fast:
@@ -4619,9 +4626,11 @@ define <3 x half> @v_log2_v3f16(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v3f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v3f16:
@@ -4726,9 +4735,11 @@ define <3 x half> @v_log2_v3f16_fast(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v3f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v3f16_fast:
@@ -4874,10 +4885,13 @@ define <4 x half> @v_log2_v4f16(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v4f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v3.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v4f16:
@@ -5026,10 +5040,13 @@ define <4 x half> @v_log2_v4f16_fast(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v4f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v3.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v4f16_fast:
diff --git a/llvm/test/CodeGen/AMDGPU/roundeven.ll b/llvm/test/CodeGen/AMDGPU/roundeven.ll
index 8920bfbd3b9dc..fd083d6eca49c 100644
--- a/llvm/test/CodeGen/AMDGPU/roundeven.ll
+++ b/llvm/test/CodeGen/AMDGPU/roundeven.ll
@@ -482,8 +482,9 @@ define <2 x half> @v_roundeven_v2f16(<2 x half> %x) {
; GFX11-TRUE16-LABEL: v_roundeven_v2f16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v1.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_roundeven_v2f16:
@@ -628,8 +629,9 @@ define <2 x half> @v_roundeven_v2f16_fneg(<2 x half> %x) {
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v1.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_roundeven_v2f16_fneg:
@@ -799,10 +801,12 @@ define <4 x half> @v_roundeven_v4f16(<4 x half> %x) {
; GFX11-TRUE16-LABEL: v_roundeven_v4f16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v0.h
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v1.l, v1.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v1.h, v1.h
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v2.l
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v1.h, v3.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_roundeven_v4f16:
diff --git a/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll b/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll
index 0787e0d0551fc..b2d9965bf1fed 100644
--- a/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll
@@ -586,9 +586,11 @@ define <3 x half> @v_constained_fsub_v3f16_fpexcept_strict(<3 x half> %x, <3 x h
; GFX11-GISEL-TRUE16-LABEL: v_constained_fsub_v3f16_fpexcept_strict:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.l, v0.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.l, v1.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v4.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_constained_fsub_v3f16_fpexcept_strict:
@@ -764,10 +766,14 @@ define <4 x half> @v_constained_fsub_v4f16_fpexcept_strict(<4 x half> %x, <4 x h
; GFX11-GISEL-TRUE16-LABEL: v_constained_fsub_v4f16_fpexcept_strict:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.l, v0.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.l, v1.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.h, v1.h, v3.h
+; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v4.l, v6.l
+; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.h, v5.l, v7.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_constained_fsub_v4f16_fpexcept_strict:
diff --git a/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll b/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll
index fd313a76fc675..fba9d6c0e6488 100644
--- a/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll
@@ -178,12 +178,13 @@ define <2 x half> @test_ldexp_v2f16_v2i32(ptr addrspace(1) %out, <2 x half> %a,
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v0, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v3, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v4, v0
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v3, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v0
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v1.l, v4.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v2f16_v2i32:
@@ -299,14 +300,15 @@ define <3 x half> @test_ldexp_v3f16_v3i32(ptr addrspace(1) %out, <3 x half> %a,
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v0, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v4, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v5, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v6, v0
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v5, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v6, 0xffff8000, v6, v0
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v4.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v1.l, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v6.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v3f16_v3i32:
@@ -445,17 +447,19 @@ define <4 x half> @test_ldexp_v4f16_v4i32(ptr addrspace(1) %out, <4 x half> %a,
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v0, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v4, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v5, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v6, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v6, 0xffff8000, v7, v0
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v3
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v5, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v6, 0xffff8000, v6, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v7, 0xffff8000, v7, v0
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v4.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v1.l, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v6.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v3.h, v6.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v8.l, v7.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v4f16_v4i32:
diff --git a/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll b/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll
index 99b6ab7a6401b..680ae989e6e21 100644
--- a/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll
@@ -1615,9 +1615,10 @@ define i16 @basic_smax_smin_vec_input(<2 x i16> %src) {
; GISEL-GFX11-TRUE16-NEXT: v_pk_min_i16 v0, 0xff00ff, v0
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX11-TRUE16-NEXT: v_pk_max_i16 v1, 0, v0
-; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
-; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GISEL-GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
; GISEL-GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GISEL-GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
@@ -1648,9 +1649,10 @@ define i16 @basic_smax_smin_vec_input(<2 x i16> %src) {
; GISEL-GFX12-TRUE16-NEXT: v_pk_min_i16 v0, 0xff00ff, v0
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX12-TRUE16-NEXT: v_pk_max_i16 v1, 0, v0
-; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
-; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GISEL-GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
; GISEL-GFX12-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GISEL-GFX12-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
@@ -1788,9 +1790,13 @@ define i16 @basic_smax_smin_vec_input_rev(<2 x i16> %src) {
; GISEL-GFX11-TRUE16-NEXT: v_pk_max_i16 v0, 0, v0
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX11-TRUE16-NEXT: v_pk_min_i16 v1, 0xff00ff, v0
-; GISEL-GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
+; GISEL-GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
+; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
+; GISEL-GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GISEL-GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
+; GISEL-GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
; GISEL-GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX11-FAKE16-LABEL: basic_smax_smin_vec_input_rev:
@@ -1818,9 +1824,13 @@ define i16 @basic_smax_smin_vec_input_rev(<2 x i16> %src) {
; GISEL-GFX12-TRUE16-NEXT: v_pk_max_i16 v0, 0, v0
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX12-TRUE16-NEXT: v_pk_min_i16 v1, 0xff00ff, v0
-; GISEL-GFX12-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
+; GISEL-GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
+; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
+; GISEL-GFX12-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GISEL-GFX12-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
+; GISEL-GFX12-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
; GISEL-GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX12-FAKE16-LABEL: basic_smax_smin_vec_input_rev:
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
index 0ec4c18a070fe..c86c1a32e4cf7 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
@@ -239,9 +239,10 @@ define half @test_vector_reduce_fadd_v3half(half %sp, <3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fadd_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -289,9 +290,10 @@ define half @test_vector_reduce_fadd_v3half(half %sp, <3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v2.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll
index 44b8b8bcb9ae8..f9fc6e9b1eea3 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll
@@ -107,9 +107,10 @@ define half @test_vector_reduce_fmax_v2half(<2 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v2half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -157,9 +158,10 @@ define half @test_vector_reduce_fmax_v2half(<2 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -286,11 +288,13 @@ define half @test_vector_reduce_fmax_v3half(<3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v2.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -346,11 +350,13 @@ define half @test_vector_reduce_fmax_v3half(<3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v2.l, v2.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -507,13 +513,16 @@ define half @test_vector_reduce_fmax_v4half(<4 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v4half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -572,13 +581,16 @@ define half @test_vector_reduce_fmax_v4half(<4 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -826,16 +838,20 @@ define half @test_vector_reduce_fmax_v8half(<8 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v8half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v4.l, v4.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v5.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v6.l, v6.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v3.h, v3.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v7.l, v7.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.h
@@ -918,16 +934,20 @@ define half @test_vector_reduce_fmax_v8half(<8 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v4.l, v4.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v5.l, v5.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v6.l, v6.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v3.h, v3.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v7.l, v7.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.h
@@ -1356,26 +1376,33 @@ define half @test_vector_reduce_fmax_v16half(<16 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v16half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v8.l, v8.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v9.l, v9.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v10.l, v10.l
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v3.h, v3.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v11.l, v11.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v4.h, v4.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v12.l, v12.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v5.l, v5.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v5.h, v5.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v13.l, v13.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.h, v6.l, v6.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v6.h, v6.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v8.l, v8.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.h, v7.l, v7.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v7.h, v7.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v9.l, v9.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v2.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.h, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v3.h, v4.l
@@ -1500,26 +1527,33 @@ define half @test_vector_reduce_fmax_v16half(<16 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v8.l, v8.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v9.l, v9.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v10.l, v10.l
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v3.h, v3.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v11.l, v11.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v4.h, v4.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v12.l, v12.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v5.l, v5.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v5.h, v5.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v13.l, v13.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v6.l, v6.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v6.h, v6.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v8.l, v8.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.h, v7.l, v7.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v7.h, v7.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v9.l, v9.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v2.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.h, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v3.h, v4.l
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll
index ed5c910def3d6..4db8e38f94b2c 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll
@@ -107,9 +107,10 @@ define half @test_vector_reduce_fmin_v2half(<2 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v2half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -157,9 +158,10 @@ define half @test_vector_reduce_fmin_v2half(<2 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -286,11 +288,13 @@ define half @test_vector_reduce_fmin_v3half(<3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v2.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -346,11 +350,13 @@ define half @test_vector_reduce_fmin_v3half(<3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v2.l, v2.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -507,13 +513,16 @@ define half @test_vector_reduce_fmin_v4half(<4 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v4half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v1.l, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v0.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -572,13 +581,16 @@ define half @test_vector_reduce_fmin_v4half(<4 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v1.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v0.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -826,16 +838,20 @@ define half @test_vector_reduce_fmin_v8half(<8 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v8half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v4.l, v4.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v5.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v6.l, v6.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v3.h, v3.h
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v1.l, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v7.l, v7.l
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v0.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.l, v2.l, v2.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.h, v3.l, v3.h
@@ -918,16 +934,20 @@ define half @test_vector_reduce_fmin_v8half(<8 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v4.l, v4.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v5.l, v5.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v6.l, v6.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v3.h, v3.h
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v1.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v7.l, v7.l
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v0.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v2.l, v2.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.h, v3.l, v3.h
@@ -1356,26 +1376,33 @@ define half @test_vector_reduce_fmin_v16half(<16 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v16half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v8.l, v8.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v9.l, v9.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v10.l, v10.l
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v1.l, v1.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.l, v2.l, v2.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v3.h, v3.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v11.l, v11.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v4.h, v4.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v12.l, v12.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v5.l, v5.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v5.h, v5.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v13.l, v13.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.h, v6.l, v6.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v6.h, v6.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v8.l, v8.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.h, v7.l, v7.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v7.h, v7.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v9.l, v9.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.h, v1.h, v2.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v2.l, v2.h, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v2.h, v3.h, v4.l
@@ -1500,26 +1527,33 @@ define half @test_vector_reduce_fmin_v16half(<16 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v8.l, v8.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v9.l, v9.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v10.l, v10.l
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v1.l, v1.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v2.l, v2.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v3.h, v3.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v11.l, v11.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v4.h, v4.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v12.l, v12.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v5.l, v5.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v5.h, v5.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v13.l, v13.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v6.l, v6.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v6.h, v6.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v8.l, v8.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.h, v7.l, v7.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v7.h, v7.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v9.l, v9.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.h, v1.h, v2.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v2.l, v2.h, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v2.h, v3.h, v4.l
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll
index 63e42e1e8a320..f85035fb3ac21 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll
@@ -100,7 +100,9 @@ define half @test_vector_reduce_fminimum_v2half(<2 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-FAKE16-LABEL: test_vector_reduce_fminimum_v2half:
@@ -241,8 +243,9 @@ define half @test_vector_reduce_fminimum_v3half(<3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v2.l
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -407,8 +410,11 @@ define half @test_vector_reduce_fminimum_v4half(<4 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v2.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v3.l
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -675,10 +681,16 @@ define half @test_vector_reduce_fminimum_v8half(<8 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v2.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v3.h
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v4.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v5.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v6.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v7.l
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
@@ -1169,14 +1181,22 @@ define half @test_vector_reduce_fminimum_v16half(<16 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v2.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v3.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.l, v4.l, v4.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.h, v5.l, v5.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.l, v6.l, v6.h
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.h, v7.l, v7.h
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v14, 16, v6
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v15, 16, v7
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v8.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v9.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v10.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v11.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.l, v4.l, v12.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.h, v5.l, v13.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.l, v6.l, v14.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.h, v7.l, v15.l
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v2.h
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll
index 57dc288bf6dcd..6e0548c35552c 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll
@@ -239,9 +239,10 @@ define half @test_vector_reduce_fmul_v3half(half %sp, <3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmul_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -289,9 +290,10 @@ define half @test_vector_reduce_fmul_v3half(half %sp, <3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v2.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
>From 483c43dd3861cc44ec9ecb3f44ba70bae1e8e394 Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Thu, 29 Jan 2026 16:19:48 -0500
Subject: [PATCH 2/2] PR feedback and undo merge/unmerge changes
Per feedback, changes involving G_MERGE_VALUES and G_UNMERGE_VALUES will not be necessary due to another patch. Additionally implement some other feedback suggestions
---
.../AMDGPU/AMDGPUInstructionSelector.cpp | 2 +-
.../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 16 +-
llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll | 52 +-
llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll | 65 +-
.../AMDGPU/GlobalISel/insertelement.ll | 2 +-
.../AMDGPU/GlobalISel/legalize-fdiv.mir | 84 +-
.../GlobalISel/legalize-load-constant.mir | 2530 +-------
.../AMDGPU/GlobalISel/legalize-load-flat.mir | 3104 +--------
.../GlobalISel/legalize-load-global.mir | 5706 +++--------------
.../AMDGPU/GlobalISel/legalize-load-local.mir | 914 ++-
.../GlobalISel/legalize-load-private.mir | 910 ++-
.../GlobalISel/legalize-store-global.mir | 1138 +---
.../test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll | 284 +-
.../test/CodeGen/AMDGPU/GlobalISel/usubsat.ll | 284 +-
.../build-vector-packed-partial-undef.ll | 54 +-
.../CodeGen/AMDGPU/dagcombine-fmul-sel.ll | 10 +-
llvm/test/CodeGen/AMDGPU/fma.f16.ll | 20 +-
llvm/test/CodeGen/AMDGPU/fmed3.ll | 23 +-
llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll | 45 +-
.../AMDGPU/llvm.amdgcn.intersect_ray.ll | 121 +-
llvm/test/CodeGen/AMDGPU/llvm.frexp.ll | 44 +-
.../CodeGen/AMDGPU/llvm.is.fpclass.f16.ll | 37 +-
llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll | 37 +-
llvm/test/CodeGen/AMDGPU/llvm.log.ll | 67 +-
llvm/test/CodeGen/AMDGPU/llvm.log10.ll | 67 +-
llvm/test/CodeGen/AMDGPU/llvm.log2.ll | 45 +-
llvm/test/CodeGen/AMDGPU/roundeven.ll | 12 +-
llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll | 12 +-
llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll | 46 +-
llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll | 26 +-
.../test/CodeGen/AMDGPU/vector-reduce-fadd.ll | 6 +-
.../test/CodeGen/AMDGPU/vector-reduce-fmax.ll | 142 +-
.../test/CodeGen/AMDGPU/vector-reduce-fmin.ll | 142 +-
.../CodeGen/AMDGPU/vector-reduce-fminimum.ll | 54 +-
.../test/CodeGen/AMDGPU/vector-reduce-fmul.ll | 6 +-
35 files changed, 3481 insertions(+), 12626 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
index 41f8773ff0e7e..e239e6f56cb44 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -665,7 +665,7 @@ bool AMDGPUInstructionSelector::selectG_MERGE_VALUES(MachineInstr &MI) const {
LLT SrcTy = MRI->getType(MI.getOperand(1).getReg());
const unsigned SrcSize = SrcTy.getSizeInBits();
- if (SrcSize < 16)
+ if (SrcSize < 32)
return selectImpl(MI, *CoverageInfo);
const DebugLoc &DL = MI.getDebugLoc();
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index b30c676622768..cf94ef7a8ef84 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -462,14 +462,8 @@ static bool isLoadStoreSizeLegal(const GCNSubtarget &ST,
MemSize = std::max(MemSize, Align);
#endif
- // Only 8-bit and 16-bit to 32-bit extloads are valid.
- if (!ST.useRealTrue16Insts() && MemSize != RegSize && RegSize != 32)
- return false;
-
- // 8-bit to 16-bit extloads are valid using True16 instructions.
- // 8-bit and 16-bit to 32-bit extloads are valid for non-True16 D16 memory
- // ops, which could be re-implemented as True16
- if (ST.useRealTrue16Insts() && MemSize != RegSize && RegSize > 32)
+ // Only allow extloads to up to 32 bits.
+ if (MemSize != RegSize && RegSize > 33)
return false;
if (MemSize > maxSizeForAddrSpace(ST, AS, IsLoad,
@@ -2038,6 +2032,7 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
auto &Builder =
getActionDefinitionsBuilder(Op)
+ .legalIf(all(isRegisterType(ST, 0), isRegisterType(ST, 1)))
.lowerFor({{S16, V2S16}})
.lowerIf([=](const LegalityQuery &Query) {
const LLT BigTy = Query.Types[BigTyIdx];
@@ -2068,10 +2063,7 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
return notValidElt(Query, BigTyIdx);
},
scalarize(1))
- .clampScalar(BigTyIdx, S32, MaxScalar)
- // legalIf must come after other rules because we don't have 16-bit SGPRs.
- // We expect 16-bit values to already be lowered or widened by this point.
- .legalIf(all(isRegisterType(ST, 0), isRegisterType(ST, 1)));
+ .clampScalar(BigTyIdx, S32, MaxScalar);
if (Op == G_MERGE_VALUES) {
Builder.widenScalarIf(
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll
index c6096e5249364..990153ebce243 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpext.ll
@@ -137,20 +137,34 @@ define amdgpu_ps void @fpext_f16_to_f64_div(half %a, ptr addrspace(1) %ptr) {
}
define amdgpu_ps <2 x float> @fpext_v2f16_to_v2f32_uniform(<2 x half> inreg %a) {
-; GFX11-LABEL: fpext_v2f16_to_v2f32_uniform:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_lshr_b32 s1, s0, 16
-; GFX11-NEXT: v_cvt_f32_f16_e32 v0, s0
-; GFX11-NEXT: v_cvt_f32_f16_e32 v1, s1
-; GFX11-NEXT: ; return to shader part epilog
+; GFX11-FAKE16-LABEL: fpext_v2f16_to_v2f32_uniform:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 16
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v0, s0
+; GFX11-FAKE16-NEXT: v_cvt_f32_f16_e32 v1, s1
+; GFX11-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX12-LABEL: fpext_v2f16_to_v2f32_uniform:
-; GFX12: ; %bb.0:
-; GFX12-NEXT: s_cvt_f32_f16 s1, s0
-; GFX12-NEXT: s_cvt_hi_f32_f16 s0, s0
-; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
-; GFX12-NEXT: v_dual_mov_b32 v0, s1 :: v_dual_mov_b32 v1, s0
-; GFX12-NEXT: ; return to shader part epilog
+; GFX11-TRUE16-LABEL: fpext_v2f16_to_v2f32_uniform:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v0
+; GFX11-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX12-FAKE16-LABEL: fpext_v2f16_to_v2f32_uniform:
+; GFX12-FAKE16: ; %bb.0:
+; GFX12-FAKE16-NEXT: s_cvt_f32_f16 s1, s0
+; GFX12-FAKE16-NEXT: s_cvt_hi_f32_f16 s0, s0
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v0, s1 :: v_dual_mov_b32 v1, s0
+; GFX12-FAKE16-NEXT: ; return to shader part epilog
+;
+; GFX12-TRUE16-LABEL: fpext_v2f16_to_v2f32_uniform:
+; GFX12-TRUE16: ; %bb.0:
+; GFX12-TRUE16-NEXT: s_cvt_f32_f16 s0, s0
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s0
+; GFX12-TRUE16-NEXT: ; return to shader part epilog
%result = fpext <2 x half> %a to <2 x float>
ret <2 x float> %result
}
@@ -166,10 +180,10 @@ define amdgpu_ps <2 x float> @fpext_v2f16_to_v2f32_div(<2 x half> %a) {
;
; GFX11-TRUE16-LABEL: fpext_v2f16_to_v2f32_div:
; GFX11-TRUE16: ; %bb.0:
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v0.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v0, v2
; GFX11-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-FAKE16-LABEL: fpext_v2f16_to_v2f32_div:
@@ -182,10 +196,10 @@ define amdgpu_ps <2 x float> @fpext_v2f16_to_v2f32_div(<2 x half> %a) {
;
; GFX12-TRUE16-LABEL: fpext_v2f16_to_v2f32_div:
; GFX12-TRUE16: ; %bb.0:
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.l
+; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v0.l
+; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v0.h
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v0, v2
; GFX12-TRUE16-NEXT: ; return to shader part epilog
%result = fpext <2 x half> %a to <2 x float>
ret <2 x float> %result
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
index c3c0ebaaf69d4..b19216ae7bbc5 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fpow.ll
@@ -496,21 +496,18 @@ define <2 x half> @v_pow_v2f16(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16-LABEL: v_pow_v2f16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
@@ -643,21 +640,19 @@ define <2 x half> @v_pow_v2f16_fneg_lhs(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
@@ -792,22 +787,20 @@ define <2 x half> @v_pow_v2f16_fneg_rhs(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16-LABEL: v_pow_v2f16_fneg_rhs:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
@@ -947,22 +940,20 @@ define <2 x half> @v_pow_v2f16_fneg_lhs_rhs(<2 x half> %x, <2 x half> %y) {
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.l
+; GFX11-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v1.l
+; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v1, v1.h
; GFX11-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v3, v0.l
-; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v2, v2.l
; GFX11-TRUE16-NEXT: v_cvt_f32_f16_e32 v0, v0.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v1, v3, v1 :: v_dual_mul_dx9_zero_f32 v2, v0, v2
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v1
+; GFX11-TRUE16-NEXT: v_dual_mul_dx9_zero_f32 v2, v3, v2 :: v_dual_mul_dx9_zero_f32 v1, v0, v1
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.l, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v2
+; GFX11-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.l, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_exp_f16_e32 v0.h, v0.h
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
index 6901d2240daed..6738c5e224267 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.ll
@@ -6533,7 +6533,7 @@ define void @insert_very_small_from_very_large(<32 x i16> %L3, ptr %ptr) {
; GFX11-LABEL: insert_very_small_from_very_large:
; GFX11: ; %bb.0: ; %bb
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v0, 1, v0
+; GFX11-NEXT: v_lshrrev_b16 v0.l, 1, v0.l
; GFX11-NEXT: v_and_b16 v0.l, v0.l, 1
; GFX11-NEXT: v_lshlrev_b16 v0.l, 1, v0.l
; GFX11-NEXT: v_and_b32_e32 v0, 3, v0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir
index 25a62320299ba..30b716ad77311 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir
@@ -1929,17 +1929,10 @@ body: |
; GFX11-NEXT: {{ $}}
; GFX11-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
- ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
- ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<2 x s16>)
+ ; GFX11-NEXT: [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<2 x s16>)
+ ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
+ ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
; GFX11-NEXT: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT1]]
; GFX11-NEXT: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32)
; GFX11-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]]
@@ -1947,13 +1940,13 @@ body: |
; GFX11-NEXT: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT]], [[FMUL]]
; GFX11-NEXT: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA1]], [[FPEXT]]
; GFX11-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FMA2]], [[INT]]
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C1]]
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C]]
; GFX11-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[AND]], [[FMA1]]
; GFX11-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32)
- ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC2]](s16), [[TRUNC]](s16)
- ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
- ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16)
+ ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[UV2]](s16), [[UV]](s16)
+ ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
+ ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
; GFX11-NEXT: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT3]]
; GFX11-NEXT: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32)
; GFX11-NEXT: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]]
@@ -1961,10 +1954,10 @@ body: |
; GFX11-NEXT: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FMA3]], [[INT2]], [[FMUL2]]
; GFX11-NEXT: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA4]], [[FPEXT2]]
; GFX11-NEXT: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FMA5]], [[INT2]]
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C1]]
+ ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C]]
; GFX11-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[AND1]], [[FMA4]]
; GFX11-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32)
- ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC3]](s16), [[TRUNC1]](s16)
+ ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[UV3]](s16), [[UV1]](s16)
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[INT1]](s16), [[INT3]](s16)
; GFX11-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
@@ -2667,27 +2660,10 @@ body: |
; GFX11-NEXT: {{ $}}
; GFX11-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
; GFX11-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
- ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
- ; GFX11-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
- ; GFX11-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
- ; GFX11-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
- ; GFX11-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
- ; GFX11-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX11-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
- ; GFX11-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
- ; GFX11-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
- ; GFX11-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
- ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)
- ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16)
+ ; GFX11-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+ ; GFX11-NEXT: [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+ ; GFX11-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16)
+ ; GFX11-NEXT: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV4]](s16)
; GFX11-NEXT: [[FNEG:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT1]]
; GFX11-NEXT: [[INT:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT1]](s32)
; GFX11-NEXT: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[INT]]
@@ -2695,13 +2671,13 @@ body: |
; GFX11-NEXT: [[FMA1:%[0-9]+]]:_(s32) = G_FMA [[FMA]], [[INT]], [[FMUL]]
; GFX11-NEXT: [[FMA2:%[0-9]+]]:_(s32) = G_FMA [[FNEG]], [[FMA1]], [[FPEXT]]
; GFX11-NEXT: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FMA2]], [[INT]]
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C1]]
+ ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 -8388608
+ ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[FMUL1]], [[C]]
; GFX11-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[AND]], [[FMA1]]
; GFX11-NEXT: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32)
- ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[TRUNC4]](s16), [[TRUNC]](s16)
- ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16)
- ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16)
+ ; GFX11-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC]](s16), [[UV4]](s16), [[UV]](s16)
+ ; GFX11-NEXT: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16)
+ ; GFX11-NEXT: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV5]](s16)
; GFX11-NEXT: [[FNEG1:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT3]]
; GFX11-NEXT: [[INT2:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT3]](s32)
; GFX11-NEXT: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[INT2]]
@@ -2709,12 +2685,12 @@ body: |
; GFX11-NEXT: [[FMA4:%[0-9]+]]:_(s32) = G_FMA [[FMA3]], [[INT2]], [[FMUL2]]
; GFX11-NEXT: [[FMA5:%[0-9]+]]:_(s32) = G_FMA [[FNEG1]], [[FMA4]], [[FPEXT2]]
; GFX11-NEXT: [[FMUL3:%[0-9]+]]:_(s32) = G_FMUL [[FMA5]], [[INT2]]
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C1]]
+ ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[FMUL3]], [[C]]
; GFX11-NEXT: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[AND1]], [[FMA4]]
; GFX11-NEXT: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32)
- ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[TRUNC5]](s16), [[TRUNC1]](s16)
- ; GFX11-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16)
- ; GFX11-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC6]](s16)
+ ; GFX11-NEXT: [[INT3:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC1]](s16), [[UV5]](s16), [[UV1]](s16)
+ ; GFX11-NEXT: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16)
+ ; GFX11-NEXT: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[UV6]](s16)
; GFX11-NEXT: [[FNEG2:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT5]]
; GFX11-NEXT: [[INT4:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT5]](s32)
; GFX11-NEXT: [[FMUL4:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[INT4]]
@@ -2722,12 +2698,12 @@ body: |
; GFX11-NEXT: [[FMA7:%[0-9]+]]:_(s32) = G_FMA [[FMA6]], [[INT4]], [[FMUL4]]
; GFX11-NEXT: [[FMA8:%[0-9]+]]:_(s32) = G_FMA [[FNEG2]], [[FMA7]], [[FPEXT4]]
; GFX11-NEXT: [[FMUL5:%[0-9]+]]:_(s32) = G_FMUL [[FMA8]], [[INT4]]
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[FMUL5]], [[C1]]
+ ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[FMUL5]], [[C]]
; GFX11-NEXT: [[FADD2:%[0-9]+]]:_(s32) = G_FADD [[AND2]], [[FMA7]]
; GFX11-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD2]](s32)
- ; GFX11-NEXT: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC6]](s16), [[TRUNC2]](s16)
- ; GFX11-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16)
- ; GFX11-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC7]](s16)
+ ; GFX11-NEXT: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[UV6]](s16), [[UV2]](s16)
+ ; GFX11-NEXT: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16)
+ ; GFX11-NEXT: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[UV7]](s16)
; GFX11-NEXT: [[FNEG3:%[0-9]+]]:_(s32) = G_FNEG [[FPEXT7]]
; GFX11-NEXT: [[INT6:%[0-9]+]]:_(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[FPEXT7]](s32)
; GFX11-NEXT: [[FMUL6:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT6]], [[INT6]]
@@ -2735,10 +2711,10 @@ body: |
; GFX11-NEXT: [[FMA10:%[0-9]+]]:_(s32) = G_FMA [[FMA9]], [[INT6]], [[FMUL6]]
; GFX11-NEXT: [[FMA11:%[0-9]+]]:_(s32) = G_FMA [[FNEG3]], [[FMA10]], [[FPEXT6]]
; GFX11-NEXT: [[FMUL7:%[0-9]+]]:_(s32) = G_FMUL [[FMA11]], [[INT6]]
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[FMUL7]], [[C1]]
+ ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[FMUL7]], [[C]]
; GFX11-NEXT: [[FADD3:%[0-9]+]]:_(s32) = G_FADD [[AND3]], [[FMA10]]
; GFX11-NEXT: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD3]](s32)
- ; GFX11-NEXT: [[INT7:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC3]](s16), [[TRUNC7]](s16), [[TRUNC3]](s16)
+ ; GFX11-NEXT: [[INT7:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC3]](s16), [[UV7]](s16), [[UV3]](s16)
; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[INT1]](s16), [[INT3]](s16)
; GFX11-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[INT5]](s16), [[INT7]](s16)
; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
index 530662cabd94c..734b460c40423 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
@@ -2,7 +2,7 @@
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=bonaire -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=CI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9 %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX10 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9 %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
@@ -39,15 +39,6 @@ body: |
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s1_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s1_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -95,15 +86,6 @@ body: |
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s2_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s2_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -145,13 +127,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s8_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s8_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -191,13 +166,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s8_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s8_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -237,13 +205,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_s16_align4
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -291,13 +252,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s16_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_s16_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -363,19 +317,6 @@ body: |
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s16_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s16_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -421,13 +362,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s32_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -484,19 +418,6 @@ body: |
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s32_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -589,29 +510,6 @@ body: |
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s32_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -666,13 +564,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s24_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), align 8, addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s24_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -712,13 +603,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s24_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s24_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -776,19 +660,6 @@ body: |
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s24_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 2, align 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s24_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -870,25 +741,6 @@ body: |
; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
; GFX9-NEXT: $vgpr0 = COPY [[OR1]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_s24_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR1]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_s24_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -946,15 +798,6 @@ body: |
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
;
- ; GFX10-LABEL: name: test_load_constant_s48_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
- ;
; GFX11-LABEL: name: test_load_constant_s48_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -996,13 +839,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10-LABEL: name: test_load_constant_s64_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11-LABEL: name: test_load_constant_s64_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1041,13 +877,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10-LABEL: name: test_load_constant_s64_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p4) :: (load (s64), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11-LABEL: name: test_load_constant_s64_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1140,31 +969,6 @@ body: |
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
;
- ; GFX10-LABEL: name: test_load_constant_s64_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
- ;
; GFX11-LABEL: name: test_load_constant_s64_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1329,49 +1133,6 @@ body: |
; GFX9-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
;
- ; GFX10-LABEL: name: test_load_constant_s64_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
- ;
; GFX11-LABEL: name: test_load_constant_s64_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1449,14 +1210,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-LABEL: name: test_load_constant_s96_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-LABEL: name: test_load_constant_s96_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1499,14 +1252,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-LABEL: name: test_load_constant_s96_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-LABEL: name: test_load_constant_s96_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1549,14 +1294,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-LABEL: name: test_load_constant_s96_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-LABEL: name: test_load_constant_s96_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1662,35 +1399,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-LABEL: name: test_load_constant_s96_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-LABEL: name: test_load_constant_s96_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -1895,61 +1603,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-LABEL: name: test_load_constant_s96_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
- ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-LABEL: name: test_load_constant_s96_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2054,19 +1707,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; GFX9-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
;
- ; GFX10-LABEL: name: test_load_constant_s160_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s32) from unknown-address + 16, addrspace 4)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
- ; GFX10-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
- ;
; GFX11-LABEL: name: test_load_constant_s160_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2138,22 +1778,6 @@ body: |
; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
;
- ; GFX10-LABEL: name: test_load_constant_s224_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 4)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
- ;
; GFX11-LABEL: name: test_load_constant_s224_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2207,14 +1831,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10-LABEL: name: test_load_constant_s128_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11-LABEL: name: test_load_constant_s128_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2257,14 +1873,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10-LABEL: name: test_load_constant_s128_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11-LABEL: name: test_load_constant_s128_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2493,76 +2101,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10-LABEL: name: test_load_constant_s128_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
- ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
- ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
- ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11-LABEL: name: test_load_constant_s128_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2667,14 +2205,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
;
- ; GFX10-LABEL: name: test_load_constant_s256_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
- ;
; GFX11-LABEL: name: test_load_constant_s256_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2714,13 +2244,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; GFX10-LABEL: name: test_load_constant_p1_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; GFX11-LABEL: name: test_load_constant_p1_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2759,13 +2282,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; GFX10-LABEL: name: test_load_constant_p1_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p4) :: (load (p1), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; GFX11-LABEL: name: test_load_constant_p1_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -2915,50 +2431,6 @@ body: |
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
;
- ; GFX10-LABEL: name: test_load_constant_p1_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
- ;
; GFX11-LABEL: name: test_load_constant_p1_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3034,13 +2506,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p4) :: (load (p3), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](p3)
;
- ; GFX10-LABEL: name: test_load_constant_p3_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p4) :: (load (p3), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](p3)
- ;
; GFX11-LABEL: name: test_load_constant_p3_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3079,13 +2544,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10-LABEL: name: test_load_constant_p4_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11-LABEL: name: test_load_constant_p4_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3124,13 +2582,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10-LABEL: name: test_load_constant_p4_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p4) :: (load (p4), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11-LABEL: name: test_load_constant_p4_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3226,32 +2677,6 @@ body: |
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
- ; GFX10-LABEL: name: test_load_constant_p4_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
- ;
; GFX11-LABEL: name: test_load_constant_p4_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3420,50 +2845,6 @@ body: |
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
- ; GFX10-LABEL: name: test_load_constant_p4_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
- ;
; GFX11-LABEL: name: test_load_constant_p4_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3539,13 +2920,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p4) :: (load (p5), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
- ; GFX10-LABEL: name: test_load_constant_p5_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p4) :: (load (p5), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
; GFX11-LABEL: name: test_load_constant_p5_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3605,20 +2979,6 @@ body: |
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
; GFX9-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
- ; GFX10-LABEL: name: test_load_constant_p5_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
- ; GFX10-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
- ;
; GFX11-LABEL: name: test_load_constant_p5_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3715,30 +3075,6 @@ body: |
; GFX9-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
; GFX9-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
- ; GFX10-LABEL: name: test_load_constant_p5_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
- ; GFX10-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
- ;
; GFX11-LABEL: name: test_load_constant_p5_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -3794,13 +3130,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v2s8_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_v2s8_align4
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -3849,13 +3178,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v2s8_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_v2s8_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -3922,31 +3244,33 @@ body: |
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v2s8_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
- ; GFX11-LABEL: name: test_load_constant_v2s8_align1
- ; GFX11: liveins: $vgpr0_vgpr1
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v2s8_align1
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[OR]](s32)
+ ; GFX11-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v2s8_align1
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 1, addrspace 4)
%2:_(s16) = G_BITCAST %1
@@ -4052,65 +3376,62 @@ body: |
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v3s8_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX11-LABEL: name: test_load_constant_v3s8_align4
- ; GFX11: liveins: $vgpr0_vgpr1
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX11-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v3s8_align4
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v3s8_align4
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 4, addrspace 4)
%2:_(s24) = G_BITCAST %1
@@ -4246,85 +3567,82 @@ body: |
; GFX9-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; GFX9-NEXT: $vgpr0 = COPY [[OR4]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v3s8_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
- ; GFX11-LABEL: name: test_load_constant_v3s8_align1
- ; GFX11: liveins: $vgpr0_vgpr1
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; GFX11-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v3s8_align1
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v3s8_align1
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX11-FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 4)
%2:_(s24) = G_BITCAST %1
@@ -4359,13 +3677,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v4s8_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_v4s8_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -4423,19 +3734,6 @@ body: |
; GFX9-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v4s8_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_v4s8_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -4529,29 +3827,6 @@ body: |
; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX10-LABEL: name: test_load_constant_v4s8_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
; GFX11-LABEL: name: test_load_constant_v4s8_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -4607,13 +3882,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v8s8_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v8s8_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -4752,62 +4020,6 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>)
;
- ; GFX10-LABEL: name: test_load_constant_v16s8_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
- ; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
- ; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
- ; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[TRUNC6]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC7]](s16), [[TRUNC8]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC9:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
- ; GFX10-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
- ; GFX10-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
- ; GFX10-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC14:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS2]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
- ; GFX10-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
- ; GFX10-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
- ; GFX10-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC15]](s16), [[TRUNC16]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC17]](s16), [[TRUNC18]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC19:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
- ; GFX10-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<16 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>)
- ;
; GFX11-LABEL: name: test_load_constant_v16s8_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -5078,106 +4290,6 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS8:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>), [[TRUNC24]](<4 x s8>), [[TRUNC29]](<4 x s8>), [[TRUNC34]](<4 x s8>), [[TRUNC39]](<4 x s8>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>)
;
- ; GFX10-LABEL: name: test_load_constant_v32s8_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
- ; GFX10-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
- ; GFX10-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
- ; GFX10-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC5]](s16), [[TRUNC6]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC7]](s16), [[TRUNC8]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC9:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS1]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
- ; GFX10-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
- ; GFX10-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
- ; GFX10-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC14:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS2]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
- ; GFX10-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
- ; GFX10-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
- ; GFX10-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC15]](s16), [[TRUNC16]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC17]](s16), [[TRUNC18]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC19:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS3]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
- ; GFX10-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
- ; GFX10-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
- ; GFX10-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC24:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS4]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
- ; GFX10-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
- ; GFX10-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
- ; GFX10-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC25]](s16), [[TRUNC26]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC27]](s16), [[TRUNC28]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS5:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC29:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS5]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
- ; GFX10-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
- ; GFX10-NEXT: [[TRUNC32:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
- ; GFX10-NEXT: [[TRUNC33:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC32]](s16), [[TRUNC33]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS6:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC34:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS6]](<4 x s16>)
- ; GFX10-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C]](s32)
- ; GFX10-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C1]](s32)
- ; GFX10-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C2]](s32)
- ; GFX10-NEXT: [[TRUNC35:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
- ; GFX10-NEXT: [[TRUNC36:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
- ; GFX10-NEXT: [[TRUNC37:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
- ; GFX10-NEXT: [[TRUNC38:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC35]](s16), [[TRUNC36]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC37]](s16), [[TRUNC38]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS7:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
- ; GFX10-NEXT: [[TRUNC39:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS7]](<4 x s16>)
- ; GFX10-NEXT: [[CONCAT_VECTORS8:%[0-9]+]]:_(<32 x s8>) = G_CONCAT_VECTORS [[TRUNC4]](<4 x s8>), [[TRUNC9]](<4 x s8>), [[TRUNC14]](<4 x s8>), [[TRUNC19]](<4 x s8>), [[TRUNC24]](<4 x s8>), [[TRUNC29]](<4 x s8>), [[TRUNC34]](<4 x s8>), [[TRUNC39]](<4 x s8>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>)
- ;
; GFX11-LABEL: name: test_load_constant_v32s8_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -5310,13 +4422,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p4) :: (load (<2 x s16>), addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p4) :: (load (<2 x s16>), addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s16_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -5381,19 +4486,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s16_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_v2s16_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -5505,29 +4597,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s16_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s16_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -5600,27 +4669,29 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s16_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX11-LABEL: name: test_load_constant_v3s16_align8
- ; GFX11: liveins: $vgpr0_vgpr1
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
- ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_constant_v3s16_align8
+ ; GFX11-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
+ ; GFX11-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_constant_v3s16_align8
+ ; GFX11-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
+ ; GFX11-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 4)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -5713,27 +4784,6 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, align 4, addrspace 4)
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_v3s16_align4
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -5864,27 +4914,6 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s16_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_v3s16_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -6057,41 +5086,6 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s16_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11-LABEL: name: test_load_constant_v3s16_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6160,13 +5154,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s16_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s16_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6205,13 +5192,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s16_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6310,29 +5290,6 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s16_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
; GFX11-TRUE16-LABEL: name: test_load_constant_v4s16_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -6520,47 +5477,6 @@ body: |
; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s16_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s16_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6636,14 +5552,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
;
- ; GFX10-LABEL: name: test_load_constant_v8s16_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
- ;
; GFX11-LABEL: name: test_load_constant_v8s16_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6683,13 +5591,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s32_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6728,13 +5629,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s32_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6815,27 +5709,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s32_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -6984,45 +5857,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s32_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7093,13 +5927,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v3s32_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7140,13 +5967,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v3s32_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7185,13 +6005,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s32_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7230,13 +6043,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s32_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7275,13 +6081,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s32_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7320,13 +6119,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v8s32_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v8s32_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7365,13 +6157,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v16s32_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v16s32_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7410,13 +6195,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s16>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
;
- ; GFX10-LABEL: name: test_load_constant_v16s32_align32_extload_from_v16s16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s16>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
- ;
; GFX11-LABEL: name: test_load_constant_v16s32_align32_extload_from_v16s16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7455,13 +6233,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s64_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s64_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7500,13 +6271,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s64_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s64_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7545,13 +6309,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s64_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s64_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -7701,50 +6458,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s64_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s16) from unknown-address + 12, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s16) from unknown-address + 14, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s64_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -8033,84 +6746,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s64_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
- ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
- ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
- ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
- ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s64_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -8220,13 +6855,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s64_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v3s64_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -8285,19 +6913,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s64_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p4) :: (load (s64) from unknown-address + 16, addrspace 4)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v3s64_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -8662,119 +7277,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v3s64_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
- ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
- ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
- ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
- ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; GFX10-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 4)
- ; GFX10-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; GFX10-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
- ; GFX10-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 4)
- ; GFX10-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
- ; GFX10-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
- ; GFX10-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
- ; GFX10-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
- ; GFX10-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 4)
- ; GFX10-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; GFX10-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
- ; GFX10-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 4)
- ; GFX10-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
- ; GFX10-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
- ; GFX10-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
- ; GFX10-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
- ; GFX10-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v3s64_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -8921,13 +7423,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s64_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s64_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -8966,13 +7461,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), align 8, addrspace 4)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s64_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), align 8, addrspace 4)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s64_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -9428,152 +7916,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s64_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
- ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
- ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
- ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
- ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; GFX10-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; GFX10-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 4)
- ; GFX10-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; GFX10-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
- ; GFX10-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 4)
- ; GFX10-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
- ; GFX10-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
- ; GFX10-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
- ; GFX10-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
- ; GFX10-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 4)
- ; GFX10-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; GFX10-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
- ; GFX10-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 4)
- ; GFX10-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
- ; GFX10-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
- ; GFX10-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
- ; GFX10-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
- ; GFX10-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
- ; GFX10-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
- ; GFX10-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C8]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p4) :: (load (s8) from unknown-address + 24, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p4) :: (load (s8) from unknown-address + 25, addrspace 4)
- ; GFX10-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
- ; GFX10-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
- ; GFX10-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p4) :: (load (s8) from unknown-address + 26, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p4) :: (load (s8) from unknown-address + 27, addrspace 4)
- ; GFX10-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
- ; GFX10-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
- ; GFX10-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
- ; GFX10-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
- ; GFX10-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
- ; GFX10-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p4) :: (load (s8) from unknown-address + 28, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p4) :: (load (s8) from unknown-address + 29, addrspace 4)
- ; GFX10-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
- ; GFX10-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
- ; GFX10-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p4) :: (load (s8) from unknown-address + 30, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p4) :: (load (s8) from unknown-address + 31, addrspace 4)
- ; GFX10-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
- ; GFX10-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
- ; GFX10-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
- ; GFX10-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
- ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
- ; GFX10-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s64_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -9754,14 +8096,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
;
- ; GFX10-LABEL: name: test_load_constant_v2s128_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
- ;
; GFX11-LABEL: name: test_load_constant_v2s128_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -9804,14 +8138,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-LABEL: name: test_load_constant_v2p1_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-LABEL: name: test_load_constant_v2p1_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -9854,14 +8180,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-LABEL: name: test_load_constant_v2p1_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-LABEL: name: test_load_constant_v2p1_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -9904,14 +8222,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-LABEL: name: test_load_constant_v2p1_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-LABEL: name: test_load_constant_v2p1_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10140,76 +8450,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-LABEL: name: test_load_constant_v2p1_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 4)
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 4)
- ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 4)
- ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 4)
- ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-LABEL: name: test_load_constant_v2p1_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10314,14 +8554,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10-LABEL: name: test_load_constant_v2p3_align8
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11-LABEL: name: test_load_constant_v2p3_align8
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10364,14 +8596,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10-LABEL: name: test_load_constant_v2p3_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11-LABEL: name: test_load_constant_v2p3_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10510,46 +8734,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10-LABEL: name: test_load_constant_v2p3_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 4)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 4)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 4)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 4)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11-LABEL: name: test_load_constant_v2p3_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10621,13 +8805,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s32_from_1_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_ext_load_constant_s32_from_1_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10666,13 +8843,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
; GFX9-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s32_from_2_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
- ; GFX10-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-LABEL: name: test_ext_load_constant_s32_from_2_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10715,14 +8885,6 @@ body: |
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s64_from_1_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-LABEL: name: test_ext_load_constant_s64_from_1_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10765,14 +8927,6 @@ body: |
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s64_from_2_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-TRUE16-LABEL: name: test_ext_load_constant_s64_from_2_align4
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -10823,14 +8977,6 @@ body: |
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s64_from_4_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-LABEL: name: test_ext_load_constant_s64_from_4_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10882,17 +9028,6 @@ body: |
; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s128_from_4_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4)
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
- ; GFX10-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
- ;
; GFX11-LABEL: name: test_ext_load_constant_s128_from_4_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -10938,14 +9073,6 @@ body: |
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s64_from_2_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4)
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-TRUE16-LABEL: name: test_ext_load_constant_s64_from_2_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-TRUE16-NEXT: {{ $}}
@@ -10996,14 +9123,6 @@ body: |
; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-LABEL: name: test_ext_load_constant_s64_from_1_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s8), align 4, addrspace 4)
- ; GFX10-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-LABEL: name: test_ext_load_constant_s64_from_1_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -11139,45 +9258,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_extload_constant_v2s32_from_4_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_extload_constant_v2s32_from_4_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -11290,27 +9370,6 @@ body: |
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_extload_constant_v2s32_from_4_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_extload_constant_v2s32_from_4_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -11363,13 +9422,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 1)
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-LABEL: name: test_extload_constant_v2s32_from_4_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-LABEL: name: test_extload_constant_v2s32_from_4_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -11408,13 +9460,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10-LABEL: name: test_extload_constant_v3s32_from_6_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11-LABEL: name: test_extload_constant_v3s32_from_6_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -11453,13 +9498,6 @@ body: |
; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-LABEL: name: test_extload_constant_v4s32_from_8_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-LABEL: name: test_extload_constant_v4s32_from_8_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -11786,109 +9824,6 @@ body: |
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align1
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p4) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p4) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p4) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p4) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; GFX10-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p4) :: (load (s8) from unknown-address + 14, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p4) :: (load (s8) from unknown-address + 15, addrspace 1)
- ; GFX10-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; GFX10-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; GFX10-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; GFX10-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p4) :: (load (s8) from unknown-address + 16, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p4) :: (load (s8) from unknown-address + 17, addrspace 1)
- ; GFX10-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; GFX10-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
- ; GFX10-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p4) :: (load (s8) from unknown-address + 18, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p4) :: (load (s8) from unknown-address + 19, addrspace 1)
- ; GFX10-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
- ; GFX10-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
- ; GFX10-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
- ; GFX10-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p4) :: (load (s8) from unknown-address + 20, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p4) :: (load (s8) from unknown-address + 21, addrspace 1)
- ; GFX10-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; GFX10-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
- ; GFX10-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p4) :: (load (s8) from unknown-address + 22, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p4) :: (load (s8) from unknown-address + 23, addrspace 1)
- ; GFX10-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
- ; GFX10-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
- ; GFX10-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
- ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align1
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -12164,59 +10099,6 @@ body: |
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align2
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s16), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p4) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p4) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p4) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p4) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p4) :: (load (s16) from unknown-address + 12, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p4) :: (load (s16) from unknown-address + 14, addrspace 1)
- ; GFX10-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p4) :: (load (s16) from unknown-address + 16, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p4) :: (load (s16) from unknown-address + 18, addrspace 1)
- ; GFX10-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
- ; GFX10-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s64)
- ; GFX10-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p4) :: (load (s16) from unknown-address + 20, addrspace 1)
- ; GFX10-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p4) :: (load (s16) from unknown-address + 22, addrspace 1)
- ; GFX10-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
- ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align2
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -12328,21 +10210,6 @@ body: |
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align4
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align4
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -12416,21 +10283,6 @@ body: |
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-LABEL: name: test_extload_constant_v2s96_from_24_align16
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-LABEL: name: test_extload_constant_v2s96_from_24_align16
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -12483,14 +10335,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
;
- ; GFX10-LABEL: name: test_load_constant_s512_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512)
- ;
; GFX11-LABEL: name: test_load_constant_s512_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
@@ -12533,14 +10377,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
;
- ; GFX10-LABEL: name: test_load_constant_v4s128_align32
- ; GFX10: liveins: $vgpr0_vgpr1
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>)
- ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>)
- ;
; GFX11-LABEL: name: test_load_constant_v4s128_align32
; GFX11: liveins: $vgpr0_vgpr1
; GFX11-NEXT: {{ $}}
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
index 397dcc4021919..f0887b64882f4 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
@@ -2,7 +2,7 @@
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=CI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9PLUS %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX10PLUS %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9PLUS %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11PLUS,GFX11PLUS-TRUE16 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11PLUS,GFX11PLUS-FAKE16 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
@@ -11,7 +11,7 @@
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=CI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX9PLUS %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX10PLUS %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX9PLUS %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX11PLUS %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=-real-true16 -mattr=-unaligned-access-mode -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=UNALIGNED_GFX12 %s
@@ -48,15 +48,6 @@ body: |
; GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s1_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -84,15 +75,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s1_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -149,15 +131,6 @@ body: |
; GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s2_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
- ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s2_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -185,15 +158,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s2_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s2_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -244,13 +208,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s8_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s8_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -272,13 +229,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s8_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -325,13 +275,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s8_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s8_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -353,13 +296,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s8_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -406,13 +342,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s16_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align4
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -450,13 +379,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s16_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -503,13 +425,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s16_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align2
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -547,13 +462,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s16_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -612,13 +520,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s16_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align1
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -662,19 +563,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s16_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -733,13 +621,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s32_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -761,13 +642,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -825,13 +699,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s32_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -859,19 +726,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -961,13 +815,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s32_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1005,29 +852,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1139,15 +963,6 @@ body: |
; GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s48_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
- ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s48_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1175,15 +990,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s48_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s48_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1242,13 +1048,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s64_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1270,13 +1069,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1330,13 +1122,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s64_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1358,13 +1143,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1438,13 +1216,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s64_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 2)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1484,31 +1255,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1654,13 +1400,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s64_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p0) :: (load (s64), align 1)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1718,49 +1457,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s64_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1895,14 +1591,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s96_align16
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -1927,14 +1615,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align16
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -1999,14 +1679,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s96_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2031,14 +1703,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2103,14 +1767,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s96_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2135,14 +1791,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2235,14 +1883,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s96_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 2)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2288,35 +1928,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2503,14 +2114,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s96_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 1)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s96_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2582,61 +2185,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s96_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s96_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2812,19 +2360,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; GFX9PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s160_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
- ; GFX10PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s160_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -2864,19 +2399,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s160_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s160_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -2987,22 +2509,6 @@ body: |
; GFX9PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s224_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
- ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s224_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3051,22 +2557,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s224_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s224_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3156,14 +2646,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s128_align16
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s128_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3188,14 +2670,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s128_align16
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s128_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3266,14 +2740,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s128_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s128_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3298,14 +2764,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s128_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s128_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3480,14 +2938,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s128_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 1)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s128_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3574,76 +3024,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s128_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s128_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3866,18 +3246,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s256_align32
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s256_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -3914,18 +3282,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s256_align32
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s256_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -3989,13 +3345,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p1_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p1_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4017,13 +3366,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p1_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p1_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4077,13 +3419,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p1_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4105,13 +3440,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p1_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4221,13 +3549,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p1_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p0) :: (load (p1), align 1)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4286,50 +3607,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p1_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4449,13 +3726,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p3_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p3_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4477,13 +3747,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p3_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p0) :: (load (p3))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p3)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p3_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4537,13 +3800,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p4_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4565,13 +3821,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4625,13 +3874,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p4_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4653,13 +3895,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4733,13 +3968,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p4_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 2)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -4780,32 +4008,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -4953,13 +4155,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p4_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p0) :: (load (p4), align 1)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p4_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5018,50 +4213,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p4_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p4_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5181,13 +4332,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p5_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p5_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5209,13 +4353,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p5_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p5_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5275,13 +4412,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p5_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 2)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p5_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5310,20 +4440,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p5_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p5_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5417,13 +4533,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_p5_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p0) :: (load (p5), align 1)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_p5_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -5462,30 +4571,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_p5_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_p5_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5565,13 +4650,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s8_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align4
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -5609,13 +4687,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s8_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5663,13 +4734,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s8_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align2
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -5707,13 +4771,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s8_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5773,13 +4830,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s8_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align1
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -5823,19 +4873,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s8_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -5966,95 +5003,119 @@ body: |
; GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s8_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX11PLUS-LABEL: name: test_load_flat_v3s8_align4
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11PLUS-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX11PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11PLUS-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX11PLUS-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX11PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX11PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX11PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX11PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX11PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_v3s8_align4
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX12-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s8_align4
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11PLUS-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11PLUS-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11PLUS-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11PLUS-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11PLUS-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11PLUS-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11PLUS-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s8_align4
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11PLUS-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11PLUS-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11PLUS-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11PLUS-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11PLUS-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11PLUS-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11PLUS-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v3s8_align4
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v3s8_align4
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v3s8_align4
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -6086,36 +5147,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s8_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6305,110 +5336,139 @@ body: |
; GFX9PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
; GFX9PLUS-NEXT: $vgpr0 = COPY [[OR3]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s8_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
- ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[OR3]](s32)
- ;
- ; GFX11PLUS-LABEL: name: test_load_flat_v3s8_align1
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
- ; GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX11PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11PLUS-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX11PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX11PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX11PLUS-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX11PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX11PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX11PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX11PLUS-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX11PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX11PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX11PLUS-NEXT: $vgpr0 = COPY [[OR3]](s32)
- ;
- ; GFX12-LABEL: name: test_load_flat_v3s8_align1
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX12-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s8_align1
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
+ ; GFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11PLUS-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11PLUS-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11PLUS-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11PLUS-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11PLUS-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11PLUS-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11PLUS-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11PLUS-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11PLUS-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11PLUS-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s8_align1
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
+ ; GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11PLUS-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11PLUS-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11PLUS-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11PLUS-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11PLUS-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11PLUS-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11PLUS-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11PLUS-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11PLUS-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11PLUS-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v3s8_align1
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v3s8_align1
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16), align 1)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 2)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v3s8_align1
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -6450,46 +5510,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR4]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s8_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6603,13 +5623,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s8_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s8_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -6631,13 +5644,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s8_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s8_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6696,13 +5702,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s8_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 2)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s8_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -6730,19 +5729,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s8_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s8_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6833,13 +5819,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s8_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32), align 1)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s8_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -6877,29 +5856,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s8_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s8_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -6986,13 +5942,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v8s8_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v8s8_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7014,13 +5963,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v8s8_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v8s8_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7087,13 +6029,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v16s8_align16
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v16s8_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7115,13 +6050,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v16s8_align16
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v16s8_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7216,17 +6144,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v32s8_align32
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v32s8_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7260,17 +6177,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v32s8_align32
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v32s8_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7326,13 +6232,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s16_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s16_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7354,13 +6253,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s16_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7426,13 +6318,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 2)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s16_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 2)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s16_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7460,19 +6345,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s16_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7570,13 +6442,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 1)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s16_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p0) :: (load (<2 x s16>), align 1)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s16_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -7614,29 +6479,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s16_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7758,38 +6600,53 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX11PLUS-LABEL: name: test_load_flat_v3s16_align8
- ; GFX11PLUS: liveins: $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: {{ $}}
- ; GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; GFX11PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX11PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align8
+ ; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; GFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; GFX11PLUS-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX12-LABEL: name: test_load_flat_v3s16_align8
- ; GFX12: liveins: $vgpr0_vgpr1
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s16_align8
+ ; GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; GFX11PLUS-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11PLUS-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_flat_v3s16_align8
+ ; GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; GFX12-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_flat_v3s16_align8
+ ; GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; GFX12-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
; UNALIGNED_GFX9PLUS-LABEL: name: test_load_flat_v3s16_align8
; UNALIGNED_GFX9PLUS: liveins: $vgpr0_vgpr1
@@ -7802,17 +6659,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -7922,27 +6768,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
- ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align4
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -8042,27 +6867,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8196,27 +7000,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align2
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -8316,27 +7099,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8498,27 +7260,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s16_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 1)
- ; GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2, align 1)
- ; GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 1)
- ; GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align1
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -8632,41 +7373,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s16_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8778,13 +7484,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8806,13 +7505,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8866,13 +7558,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -8894,13 +7579,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -8988,13 +7666,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 2)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9032,29 +7703,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9210,13 +7858,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s16_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>), align 1)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s16_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9272,47 +7913,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s16_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9449,14 +8049,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v8s16_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v8s16_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9481,14 +8073,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v8s16_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v8s16_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9544,13 +8128,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s32_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s32_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9572,13 +8149,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s32_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s32_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9632,13 +8202,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s32_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9660,13 +8223,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s32_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9721,13 +8277,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s32_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s32_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9749,13 +8298,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s32_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s32_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9815,13 +8357,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s32_align16
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s32_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9843,13 +8378,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s32_align16
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s32_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -9911,13 +8439,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s32_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -9939,13 +8460,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s32_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10011,13 +8525,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s32_align16
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s32_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10039,13 +8546,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s32_align16
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s32_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10111,13 +8611,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s32_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s32_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10139,13 +8632,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s32_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s32_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10211,13 +8697,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s32_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s32_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10239,13 +8718,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s32_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s32_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10339,17 +8811,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v8s32_align32
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v8s32_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10383,17 +8844,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v8s32_align32
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<8 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v8s32_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10549,23 +8999,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v16s32_align32
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
- ; GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p0) :: (load (<4 x s32>) from unknown-address + 32, align 32)
- ; GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 48
- ; GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p0) :: (load (<4 x s32>) from unknown-address + 48)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v16s32_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10617,23 +9050,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v16s32_align32
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD1]](p0) :: (load (<4 x s32>) from unknown-address + 32, align 32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 48
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD2]](p0) :: (load (<4 x s32>) from unknown-address + 48)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[CONCAT_VECTORS]](<16 x s32>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v16s32_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10721,13 +9137,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align16
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10749,13 +9158,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align16
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>))
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10823,13 +9225,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10851,13 +9246,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -10925,13 +9313,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -10953,13 +9334,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11063,13 +9437,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 2)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 2)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align2
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11128,50 +9495,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s16) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p0) :: (load (s16) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s16) from unknown-address + 12)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s16) from unknown-address + 14)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11417,13 +9740,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s64_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11516,84 +9832,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s64_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11823,19 +10061,6 @@ body: |
; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s64_align32
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
- ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s64_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -11875,19 +10100,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s64_align32
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s64_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -11989,19 +10201,6 @@ body: |
; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s64_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12041,19 +10240,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s64_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12307,19 +10493,6 @@ body: |
; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v3s64_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 1)
- ; GFX10PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v3s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12459,119 +10632,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v3s64_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -12881,17 +10941,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s64_align32
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s64_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -12925,17 +10974,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s64_align32
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s64_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -13039,17 +11077,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s64_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 8)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s64_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -13083,17 +11110,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s64_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s64_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -13397,17 +11413,6 @@ body: |
; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v4s64_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 1)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[PTR_ADD]](p0) :: (load (<2 x s64>) from unknown-address + 16, align 1)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[LOAD]](<2 x s64>), [[LOAD1]](<2 x s64>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v4s64_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -13577,153 +11582,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s64>), [[BUILD_VECTOR1]](<2 x s64>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v4s64_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p0) :: (load (s8) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p0) :: (load (s8) from unknown-address + 17)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p0) :: (load (s8) from unknown-address + 18)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p0) :: (load (s8) from unknown-address + 19)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p0) :: (load (s8) from unknown-address + 20)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p0) :: (load (s8) from unknown-address + 21)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p0) :: (load (s8) from unknown-address + 22)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p0) :: (load (s8) from unknown-address + 23)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p0) :: (load (s8) from unknown-address + 24)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p0) :: (load (s8) from unknown-address + 25)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p0) :: (load (s8) from unknown-address + 26)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p0) :: (load (s8) from unknown-address + 27)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p0) :: (load (s8) from unknown-address + 28)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p0) :: (load (s8) from unknown-address + 29)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p0) :: (load (s8) from unknown-address + 30)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p0) :: (load (s8) from unknown-address + 31)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR20]](s64), [[OR27]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s64>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s64>), [[BUILD_VECTOR1]](<2 x s64>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<4 x s64>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v4s64_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -14100,18 +11958,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2s128_align32
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2s128_align32
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -14148,18 +11994,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2s128_align32
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<4 x s32>) from unknown-address + 16)
- ; UNALIGNED_GFX10PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s128_align32
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -14238,14 +12072,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align16
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align16
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -14270,14 +12096,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align16
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align16
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -14348,14 +12166,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -14380,14 +12190,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -14458,14 +12260,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -14490,14 +12284,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -14672,14 +12458,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2p1_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 1)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -14766,76 +12544,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p1_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p0) :: (load (s8) from unknown-address + 8)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p0) :: (load (s8) from unknown-address + 9)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p0) :: (load (s8) from unknown-address + 10)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p0) :: (load (s8) from unknown-address + 11)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p0) :: (load (s8) from unknown-address + 12)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p0) :: (load (s8) from unknown-address + 13)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p0) :: (load (s8) from unknown-address + 14)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p0) :: (load (s8) from unknown-address + 15)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15018,14 +12726,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2p3_align8
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p3_align8
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15050,14 +12750,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p3_align8
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>))
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p3_align8
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15116,14 +12808,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2p3_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p3_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15148,14 +12832,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p3_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p3_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15270,14 +12946,6 @@ body: |
; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_v2p3_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>), align 1)
- ; GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_v2p3_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15334,46 +13002,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_v2p3_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX10PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX10PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s8) from unknown-address + 3)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p0) :: (load (s8) from unknown-address + 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p0) :: (load (s8) from unknown-address + 5)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p0) :: (load (s8) from unknown-address + 6)
- ; UNALIGNED_GFX10PLUS-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p0) :: (load (s8) from unknown-address + 7)
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; UNALIGNED_GFX10PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2p3_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15485,13 +13113,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15513,13 +13134,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15565,13 +13179,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15593,13 +13200,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s32_from_2_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15649,14 +13249,6 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15681,14 +13273,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15739,14 +13323,6 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -15787,14 +13363,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15845,14 +13413,6 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15877,14 +13437,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_4_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -15944,17 +13496,6 @@ body: |
; GFX9PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
- ; GFX10PLUS-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
- ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -15988,17 +13529,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX10PLUS-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s128_from_4_align4
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -16055,14 +13585,6 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11PLUS-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
; GFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
; GFX11PLUS-TRUE16-NEXT: {{ $}}
@@ -16103,14 +13625,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_2_align2
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -16161,14 +13675,6 @@ body: |
; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -16193,14 +13699,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s8), align 4)
- ; UNALIGNED_GFX10PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_ext_load_flat_s64_from_1_align1
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
@@ -16249,13 +13747,6 @@ body: |
; GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
; GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10PLUS-LABEL: name: test_load_flat_s32_align536870912
- ; GFX10PLUS: liveins: $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: {{ $}}
- ; GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
- ; GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11PLUS-LABEL: name: test_load_flat_s32_align536870912
; GFX11PLUS: liveins: $vgpr0_vgpr1
; GFX11PLUS-NEXT: {{ $}}
@@ -16277,13 +13768,6 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX10PLUS-LABEL: name: test_load_flat_s32_align536870912
- ; UNALIGNED_GFX10PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX10PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX10PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 536870912)
- ; UNALIGNED_GFX10PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s32_align536870912
; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
index 77d72f62b5e3a..a171ff6984712 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
@@ -12,8 +12,8 @@
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-MESA %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX10-HSA %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX10-MESA %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-HSA %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX9-MESA %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-HSA,GFX11-HSA-TRUE16 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-HSA,GFX11-HSA-FAKE16 %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-MESA,GFX11-MESA-TRUE16 %s
@@ -91,24 +91,6 @@ body: |
; GFX9-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s1_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s1_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s1_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -210,24 +192,6 @@ body: |
; GFX9-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s2_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
- ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s2_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
- ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[AND]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s2_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -317,20 +281,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s8_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s8_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s8_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -412,20 +362,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s8_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s8_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s8_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -507,20 +443,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_s16_align4
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -634,20 +556,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_s16_align2
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -785,26 +693,6 @@ body: |
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_s16_align1
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -914,20 +802,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s32_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s32_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s32_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -1032,26 +906,6 @@ body: |
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s32_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s32_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s32_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -1208,36 +1062,6 @@ body: |
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s32_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s32_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s32_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -1350,20 +1174,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s24_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s24_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s24_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -1445,20 +1255,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s24_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s24_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s24_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -1576,32 +1372,6 @@ body: |
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s24_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s24_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s24_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -1767,38 +1537,6 @@ body: |
; GFX9-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s24_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s24_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s24_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -1928,24 +1666,6 @@ body: |
; GFX9-MESA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s48_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
- ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s48_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 281474976710655
- ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[LOAD]], [[C]]
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[AND]](s64)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s48_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -2035,20 +1755,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s64_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s64_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s64_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -2129,20 +1835,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s64_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s64_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s64_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -2295,38 +1987,6 @@ body: |
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s64_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s64_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR2]](s64)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s64_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -2587,56 +2247,6 @@ body: |
; GFX9-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s64_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p1) :: (load (s64), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s64_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[OR6]](s64)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s64_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -2797,22 +2407,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s96_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s96_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s96_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -2908,22 +2502,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s96_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s96_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s96_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -3019,22 +2597,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s96_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s96_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s96_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -3209,43 +2771,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s96_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s96_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s96_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -3566,69 +3091,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s96_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s96_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s96_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -3843,32 +3305,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
; GFX9-MESA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s160_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
- ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
- ; GFX10-HSA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s160_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from unknown-address + 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[BUILD_VECTOR]](<5 x s32>)
- ; GFX10-MESA-NEXT: S_NOP 0, implicit [[BITCAST]](s160)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s160_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -4030,38 +3466,6 @@ body: |
; GFX9-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s224_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s224_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s224_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -4187,22 +3591,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s128_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s128_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s128_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -4293,22 +3681,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s128_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s128_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s128_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -4647,84 +4019,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s128_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s128_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s128_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -4939,22 +4233,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s256_align32
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s256_align32
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s256_align32
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -5039,20 +4317,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p1_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p1_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p1_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -5133,20 +4397,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p1_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p1_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p1_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -5375,57 +4625,6 @@ body: |
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p1_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[COPY]](p1) :: (load (p1), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p1_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p1) = G_INTTOPTR [[OR6]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p1)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p1_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -5580,20 +4779,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p3_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p3_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p3) = G_LOAD [[COPY]](p1) :: (load (p3), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p3)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p3_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -5674,20 +4859,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p4_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p4_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p4_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -5768,20 +4939,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p4_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p4_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p4_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -5938,39 +5095,6 @@ body: |
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p4_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p4_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR2]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p4_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -6237,57 +5361,6 @@ body: |
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p4_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p4) = G_LOAD [[COPY]](p1) :: (load (p4), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p4)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p4_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p4) = G_INTTOPTR [[OR6]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[INTTOPTR]](p4)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p4_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -6442,20 +5515,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p5_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p5_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p5_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -6564,27 +5623,6 @@ body: |
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p5_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p5_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR]](s32)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p5_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -6747,37 +5785,6 @@ body: |
; GFX9-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
;
- ; GFX10-HSA-LABEL: name: test_load_global_p5_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(p5) = G_LOAD [[COPY]](p1) :: (load (p5), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](p5)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_p5_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[INTTOPTR:%[0-9]+]]:_(p5) = G_INTTOPTR [[OR2]](s32)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[INTTOPTR]](p5)
- ;
; GFX11-HSA-LABEL: name: test_load_global_p5_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -6892,20 +5899,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s8_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s8_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align4
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -7020,20 +6013,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s8_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s8_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align2
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -7172,26 +6151,6 @@ body: |
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s8_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s8_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align1
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -7207,18 +6166,33 @@ body: |
; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX11-MESA-LABEL: name: test_load_global_v2s8_align1
- ; GFX11-MESA: liveins: $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: {{ $}}
- ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[OR]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[OR]](s32)
;
; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v2s8_align1
; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
@@ -7235,18 +6209,33 @@ body: |
; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX12-MESA-LABEL: name: test_load_global_v2s8_align1
- ; GFX12-MESA: liveins: $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: {{ $}}
- ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[OR]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 1, addrspace 1)
%2:_(s16) = G_BITCAST %1
@@ -7446,185 +6435,233 @@ body: |
; GFX9-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s8_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX10-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX10-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX10-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX10-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX10-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s8_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX11-HSA-LABEL: name: test_load_global_v3s8_align4
- ; GFX11-HSA: liveins: $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: {{ $}}
- ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX11-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX11-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX11-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX11-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX11-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX11-MESA-LABEL: name: test_load_global_v3s8_align4
- ; GFX11-MESA: liveins: $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: {{ $}}
- ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX12-HSA-LABEL: name: test_load_global_v3s8_align4
- ; GFX12-HSA: liveins: $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: {{ $}}
- ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX12-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX12-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX12-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX12-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX12-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX12-MESA-LABEL: name: test_load_global_v3s8_align4
- ; GFX12-MESA: liveins: $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: {{ $}}
- ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-HSA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-HSA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-HSA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-HSA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-HSA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-HSA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-HSA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-HSA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-MESA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-MESA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-MESA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-MESA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-MESA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-HSA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-HSA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-HSA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-HSA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-HSA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-HSA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-HSA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-HSA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX-MESA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX-MESA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX-MESA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX-MESA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX-MESA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX-MESA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX-MESA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX-MESA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX-MESA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX-MESA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 4, addrspace 1)
%2:_(s24) = G_BITCAST %1
@@ -7875,230 +6912,293 @@ body: |
; GFX9-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s8_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX10-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX10-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX10-HSA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX10-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX10-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX10-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR3]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s8_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX10-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX10-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
- ; GFX11-HSA-LABEL: name: test_load_global_v3s8_align1
- ; GFX11-HSA: liveins: $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: {{ $}}
- ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX11-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX11-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX11-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX11-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX11-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX11-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX11-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX11-HSA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX11-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX11-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX11-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX11-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX11-HSA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX11-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX11-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX11-HSA-NEXT: $vgpr0 = COPY [[OR3]](s32)
- ;
- ; GFX11-MESA-LABEL: name: test_load_global_v3s8_align1
- ; GFX11-MESA: liveins: $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: {{ $}}
- ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX11-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX11-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX11-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX11-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX11-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX11-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX11-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX11-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX11-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; GFX11-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX11-MESA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; GFX11-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; GFX11-MESA-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; GFX11-MESA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; GFX11-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; GFX11-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; GFX11-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; GFX11-MESA-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; GFX11-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX11-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; GFX11-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; GFX11-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; GFX11-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
- ; GFX12-HSA-LABEL: name: test_load_global_v3s8_align1
- ; GFX12-HSA: liveins: $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: {{ $}}
- ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX12-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX12-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX12-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX12-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX12-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX12-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX12-HSA-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-HSA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX12-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-HSA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX12-HSA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX12-HSA-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX12-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-HSA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX12-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-HSA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX12-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX12-HSA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX12-HSA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-HSA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX12-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX12-HSA-NEXT: $vgpr0 = COPY [[OR3]](s32)
- ;
- ; GFX12-MESA-LABEL: name: test_load_global_v3s8_align1
- ; GFX12-MESA: liveins: $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: {{ $}}
- ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX12-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX12-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX12-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX12-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX12-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX12-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX12-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX12-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX12-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX12-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; GFX12-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX12-MESA-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-MESA-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; GFX12-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-MESA-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; GFX12-MESA-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; GFX12-MESA-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; GFX12-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-MESA-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; GFX12-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-MESA-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; GFX12-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; GFX12-MESA-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; GFX12-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX12-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; GFX12-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; GFX12-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; GFX12-MESA-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-HSA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-HSA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11-HSA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-HSA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11-HSA-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11-HSA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-HSA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-HSA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-HSA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-HSA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11-HSA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11-HSA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-HSA-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-MESA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-MESA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-MESA-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-MESA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-MESA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-MESA-TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-MESA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX11-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-MESA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-MESA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-MESA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX11-MESA-FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-HSA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-HSA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-HSA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-HSA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-HSA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-HSA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-HSA-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-HSA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-HSA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-HSA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-HSA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-HSA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-HSA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-HSA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-HSA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-HSA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-HSA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-HSA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-HSA-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX-MESA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX-MESA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX-MESA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX-MESA-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX-MESA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX-MESA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX-MESA-TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX-MESA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX-MESA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX-MESA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX-MESA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX-MESA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX-MESA-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX-MESA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX-MESA-FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 1)
%2:_(s24) = G_BITCAST %1
@@ -8214,48 +7314,6 @@ body: |
; GFX9-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s8_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s8_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s8_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -8472,53 +7530,6 @@ body: |
; GFX9-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s8_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s8_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s8_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -8781,62 +7792,6 @@ body: |
; GFX9-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s8_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s8_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR2]], [[C4]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[CONCAT_VECTORS]](<4 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC4]](<4 x s8>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s8_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -9001,20 +7956,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v8s8_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v8s8_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v8s8_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -9096,20 +8037,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v16s8_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v16s8_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v16s8_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -9191,20 +8118,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v32s8_align32
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v32s8_align32
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v32s8_align32
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -9287,20 +8200,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s16_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -9417,26 +8316,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s16_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -9627,36 +8506,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s16_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -9805,71 +8654,101 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX11-HSA-LABEL: name: test_load_global_v3s16_align8
- ; GFX11-HSA: liveins: $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: {{ $}}
- ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX11-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-MESA-LABEL: name: test_load_global_v3s16_align8
- ; GFX11-MESA: liveins: $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: {{ $}}
- ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX11-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX12-HSA-LABEL: name: test_load_global_v3s16_align8
- ; GFX12-HSA: liveins: $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: {{ $}}
- ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX12-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-HSA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX12-MESA-LABEL: name: test_load_global_v3s16_align8
- ; GFX12-MESA: liveins: $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: {{ $}}
- ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX12-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 1)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -10041,48 +8920,6 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align4
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -10409,48 +9246,6 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align2
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -10833,62 +9628,6 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v3s16_align1
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -11091,20 +9830,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s16_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -11185,20 +9910,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s16_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -11361,36 +10072,6 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s16_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -11695,54 +10376,6 @@ body: |
; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<4 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s16_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -11919,83 +10552,117 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
- ;
- ; GFX11-HSA-LABEL: name: test_load_global_v5s16_align16
- ; GFX11-HSA: liveins: $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: {{ $}}
- ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX11-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX11-HSA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
- ;
- ; GFX11-MESA-LABEL: name: test_load_global_v5s16_align16
- ; GFX11-MESA: liveins: $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: {{ $}}
- ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX11-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX11-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
- ;
- ; GFX12-HSA-LABEL: name: test_load_global_v5s16_align16
- ; GFX12-HSA: liveins: $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: {{ $}}
- ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX12-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX12-HSA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
- ;
- ; GFX12-MESA-LABEL: name: test_load_global_v5s16_align16
- ; GFX12-MESA: liveins: $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: {{ $}}
- ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX12-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX12-MESA-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV4]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV4]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV4]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV4]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX-MESA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[UV2]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<5 x s16>) = G_LOAD %0 :: (load (<5 x s16>), align 16, addrspace 1)
%2:_(<5 x s16>) = G_IMPLICIT_DEF
@@ -12207,66 +10874,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align8
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -12697,66 +11304,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align4
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -13205,66 +11752,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align2
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -13801,88 +12288,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v5s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v5s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v5s16_align1
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -14163,22 +12568,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v6s16_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -14274,22 +12663,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v6s16_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -14385,22 +12758,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v6s16_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -14575,43 +12932,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v6s16_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -14932,69 +13252,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v6s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v6s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v6s16_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -15209,95 +13466,133 @@ body: |
; GFX9-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-HSA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-MESA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
- ;
- ; GFX11-HSA-LABEL: name: test_load_global_v7s16_align16
- ; GFX11-HSA: liveins: $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: {{ $}}
- ; GFX11-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX11-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX11-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-HSA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX11-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX11-HSA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
- ; GFX11-HSA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
- ;
- ; GFX11-MESA-LABEL: name: test_load_global_v7s16_align16
- ; GFX11-MESA: liveins: $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: {{ $}}
- ; GFX11-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX11-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX11-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX11-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-MESA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX11-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX11-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX11-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
- ; GFX11-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
- ;
- ; GFX12-HSA-LABEL: name: test_load_global_v7s16_align16
- ; GFX12-HSA: liveins: $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: {{ $}}
- ; GFX12-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX12-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX12-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-HSA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-HSA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-HSA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX12-HSA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX12-HSA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
- ; GFX12-HSA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
- ;
- ; GFX12-MESA-LABEL: name: test_load_global_v7s16_align16
- ; GFX12-MESA: liveins: $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: {{ $}}
- ; GFX12-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX12-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX12-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX12-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-MESA-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-MESA-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
- ; GFX12-MESA-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
- ; GFX12-MESA-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
- ; GFX12-MESA-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
- ; GFX12-MESA-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX11-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV6]](s16), [[DEF]](s16)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr2 = COPY [[UV18]](<2 x s16>)
+ ; GFX11-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX11-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX11-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-HSA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX11-HSA-FAKE16-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX11-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX11-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX11-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV6]](s16), [[DEF]](s16)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr2 = COPY [[UV18]](<2 x s16>)
+ ; GFX11-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX11-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX11-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX11-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX11-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX11-MESA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX11-MESA-FAKE16-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX12-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX12-HSA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: {{ $}}
+ ; GFX12-HSA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-HSA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV6]](s16), [[DEF]](s16)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr2 = COPY [[UV18]](<2 x s16>)
+ ; GFX12-HSA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-HSA-FAKE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX12-HSA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: {{ $}}
+ ; GFX12-HSA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-HSA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-HSA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX12-HSA-FAKE16-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
+ ;
+ ; GFX-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX-MESA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV6]](s16), [[DEF]](s16)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr2 = COPY [[UV18]](<2 x s16>)
+ ; GFX-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX-MESA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX-MESA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX-MESA-FAKE16-NEXT: $vgpr3 = COPY [[UV3]](<2 x s16>)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(<7 x s16>) = G_LOAD %0 :: (load (<7 x s16>), align 16, addrspace 1)
%2:_(<7 x s16>) = G_IMPLICIT_DEF
@@ -15596,86 +13891,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
- ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align8
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -16265,86 +14480,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
- ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align4
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -16934,86 +15069,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
- ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align2
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -17723,116 +15778,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
; GFX9-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v7s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
- ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
- ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v7s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[OR3]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[OR4]](s32)
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[OR6]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
- ; GFX10-MESA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_load_global_v7s16_align1
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -18184,22 +16129,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v8s16_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v8s16_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v8s16_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -18290,22 +16219,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v8s16_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v8s16_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v8s16_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -18390,20 +16303,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s32_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -18484,20 +16383,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s32_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -18634,34 +16519,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s32_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -18898,52 +16755,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s32_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s32_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s32_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19090,20 +16901,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s32_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s32_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v3s32_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19189,20 +16986,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s32_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s32_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v3s32_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19283,20 +17066,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s32_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s32_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s32_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19377,20 +17146,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s32_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s32_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s32_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19471,20 +17226,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s32_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s32_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s32_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19565,20 +17306,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v8s32_align32
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v8s32_align32
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<8 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v8s32_align32
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19659,20 +17386,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v16s32_align32
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v16s32_align32
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 32, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[LOAD]](<16 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v16s32_align32
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19753,20 +17466,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s64_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19847,20 +17546,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s64_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -19941,20 +17626,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s64_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -20183,57 +17854,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR1]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s64) = G_OR [[SHL2]], [[ZEXT]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR3]](s32)
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR4]](s32)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s64) = G_OR [[SHL5]], [[ZEXT1]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR2]](s64), [[OR5]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s64_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -20672,91 +18292,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s64_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s64_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s64_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -20985,22 +18520,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2sp1_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2sp1_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2sp1_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -21085,20 +18604,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s64_align32
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s64_align32
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v3s64_align32
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -21217,32 +18722,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s64_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
- ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s64_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v3s64_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -21785,132 +19264,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v3s64_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[DEF]](s64)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v3s64_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
- ; GFX10-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
- ; GFX10-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
- ; GFX10-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
- ; GFX10-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
- ; GFX10-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
- ; GFX10-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
- ; GFX10-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
- ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
- ; GFX10-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[DEF]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v3s64_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -22217,20 +19570,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s64_align32
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s64_align32
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s64_align32
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -22311,20 +19650,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s64_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s64_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s64_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -22961,159 +20286,6 @@ body: |
; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4s64_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[LOAD]](<4 x s64>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4s64_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s64) = G_OR [[SHL6]], [[ZEXT]]
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[OR8]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[OR7]]
- ; GFX10-MESA-NEXT: [[ZEXT1:%[0-9]+]]:_(s64) = G_ZEXT [[OR9]](s32)
- ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD9]]
- ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[ZEXTLOAD11]]
- ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[OR11]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[OR10]]
- ; GFX10-MESA-NEXT: [[ANYEXT1:%[0-9]+]]:_(s64) = G_ANYEXT [[OR12]](s32)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT1]], [[COPY1]](s32)
- ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s64) = G_OR [[SHL13]], [[ZEXT1]]
- ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C7]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[ZEXTLOAD12]]
- ; GFX10-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD14]]
- ; GFX10-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[OR15]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[OR14]]
- ; GFX10-MESA-NEXT: [[ZEXT2:%[0-9]+]]:_(s64) = G_ZEXT [[OR16]](s32)
- ; GFX10-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[ZEXTLOAD15]]
- ; GFX10-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
- ; GFX10-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
- ; GFX10-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
- ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
- ; GFX10-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
- ; GFX10-MESA-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
- ; GFX10-MESA-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C8]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD18:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD23]](p1) :: (load (s8) from unknown-address + 24, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD19:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD24]](p1) :: (load (s8) from unknown-address + 25, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL21:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD19]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR21:%[0-9]+]]:_(s32) = G_OR [[SHL21]], [[ZEXTLOAD18]]
- ; GFX10-MESA-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD20:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD25]](p1) :: (load (s8) from unknown-address + 26, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD25]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD26]](p1) :: (load (s8) from unknown-address + 27, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL22:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR22:%[0-9]+]]:_(s32) = G_OR [[SHL22]], [[ZEXTLOAD20]]
- ; GFX10-MESA-NEXT: [[SHL23:%[0-9]+]]:_(s32) = G_SHL [[OR22]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR23:%[0-9]+]]:_(s32) = G_OR [[SHL23]], [[OR21]]
- ; GFX10-MESA-NEXT: [[ZEXT3:%[0-9]+]]:_(s64) = G_ZEXT [[OR23]](s32)
- ; GFX10-MESA-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD23]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD21:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD27]](p1) :: (load (s8) from unknown-address + 28, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD22:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD28]](p1) :: (load (s8) from unknown-address + 29, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL24:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD22]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR24:%[0-9]+]]:_(s32) = G_OR [[SHL24]], [[ZEXTLOAD21]]
- ; GFX10-MESA-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD27]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD23:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD29]](p1) :: (load (s8) from unknown-address + 30, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD29]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD7:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD30]](p1) :: (load (s8) from unknown-address + 31, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL25:%[0-9]+]]:_(s32) = G_SHL [[LOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR25:%[0-9]+]]:_(s32) = G_OR [[SHL25]], [[ZEXTLOAD23]]
- ; GFX10-MESA-NEXT: [[SHL26:%[0-9]+]]:_(s32) = G_SHL [[OR25]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR26:%[0-9]+]]:_(s32) = G_OR [[SHL26]], [[OR24]]
- ; GFX10-MESA-NEXT: [[ANYEXT3:%[0-9]+]]:_(s64) = G_ANYEXT [[OR26]](s32)
- ; GFX10-MESA-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
- ; GFX10-MESA-NEXT: [[SHL27:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT3]], [[COPY3]](s32)
- ; GFX10-MESA-NEXT: [[OR27:%[0-9]+]]:_(s64) = G_OR [[SHL27]], [[ZEXT3]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[OR27]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4s64_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -23478,22 +20650,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2s128_align32
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2s128_align32
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2s128_align32
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -23584,22 +20740,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2p1_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -23690,22 +20830,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2p1_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -23796,22 +20920,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2p1_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -24150,84 +21258,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2p1_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2p1_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[BUILD_VECTOR]](<4 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2p1_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -24442,22 +21472,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v4p1_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v4p1_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v4p1_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -24548,22 +21562,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2p3_align8
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2p3_align8
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2p3_align8
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -24654,22 +21652,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2p3_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2p3_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2p3_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -24888,54 +21870,6 @@ body: |
; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v2p3_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[LOAD]](<2 x s32>)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v2p3_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x p3>) = G_BITCAST [[BUILD_VECTOR]](<2 x s32>)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[BITCAST]](<2 x p3>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v2p3_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -25084,20 +22018,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_1_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_1_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_1_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -25178,20 +22098,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_2_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_2_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_2_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -25332,38 +22238,6 @@ body: |
; GFX9-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR1]](s32)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -25515,32 +22389,6 @@ body: |
; GFX9-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; GFX9-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, align 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -25645,20 +22493,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s32_from_s24_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s32_from_s24_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -25746,22 +22580,6 @@ body: |
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_1_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_1_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s64_from_1_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -25852,22 +22670,6 @@ body: |
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_2_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_2_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align4
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -25990,22 +22792,6 @@ body: |
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_4_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_4_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s64_from_4_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -26114,28 +22900,6 @@ body: |
; GFX9-MESA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s128_from_4_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
- ; GFX10-HSA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-HSA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s128_from_4_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[DEF]](s32)
- ; GFX10-MESA-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
- ; GFX10-MESA-NEXT: [[MV1:%[0-9]+]]:_(s128) = G_MERGE_VALUES [[MV]](s64), [[DEF1]](s64)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[MV1]](s128)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s128_from_4_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -26238,22 +23002,6 @@ body: |
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_2_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_2_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-HSA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align2
; GFX11-HSA-TRUE16: liveins: $vgpr0_vgpr1
; GFX11-HSA-TRUE16-NEXT: {{ $}}
@@ -26376,22 +23124,6 @@ body: |
; GFX9-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
;
- ; GFX10-HSA-LABEL: name: test_ext_load_global_s64_from_1_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; GFX10-MESA-LABEL: name: test_ext_load_global_s64_from_1_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s8), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
; GFX11-HSA-LABEL: name: test_ext_load_global_s64_from_1_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -26476,20 +23208,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 1, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -26570,20 +23288,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), align 2, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -26664,20 +23368,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s16>), addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_extload_global_v2s32_from_v2s16_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -26758,20 +23448,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s16>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_extload_global_v3s32_from_v3s16_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -26852,20 +23528,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
;
- ; GFX10-HSA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
- ; GFX10-MESA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<4 x s32>)
- ;
; GFX11-HSA-LABEL: name: test_extload_global_v4s32_from_v4s16_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -27346,124 +24008,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-HSA-LABEL: name: test_global_v2s96_align1
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_global_v2s96_align1
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s8) from unknown-address + 3, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[OR1]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[OR]]
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s8) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD4]](p1) :: (load (s8) from unknown-address + 5, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s8) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s8) from unknown-address + 7, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[OR4]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[OR3]]
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD7:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD7]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD8:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s8) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s8) from unknown-address + 11, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[SHL7]], [[ZEXTLOAD8]]
- ; GFX10-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[OR7]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[SHL8]], [[OR6]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C6]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD9:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD10:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD10]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[SHL9]], [[ZEXTLOAD9]]
- ; GFX10-MESA-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD11:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD13]](p1) :: (load (s8) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD13]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD14]](p1) :: (load (s8) from unknown-address + 15, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[SHL10]], [[ZEXTLOAD11]]
- ; GFX10-MESA-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[OR10]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[SHL11]], [[OR9]]
- ; GFX10-MESA-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD12:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD15]](p1) :: (load (s8) from unknown-address + 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD13:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD16]](p1) :: (load (s8) from unknown-address + 17, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD13]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[SHL12]], [[ZEXTLOAD12]]
- ; GFX10-MESA-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD15]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD14:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD17]](p1) :: (load (s8) from unknown-address + 18, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD17]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD18]](p1) :: (load (s8) from unknown-address + 19, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[SHL13]], [[ZEXTLOAD14]]
- ; GFX10-MESA-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[OR13]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[SHL14]], [[OR12]]
- ; GFX10-MESA-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD11]], [[C5]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD15:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD19]](p1) :: (load (s8) from unknown-address + 20, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD16:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD20]](p1) :: (load (s8) from unknown-address + 21, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD16]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[SHL15]], [[ZEXTLOAD15]]
- ; GFX10-MESA-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD19]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD17:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD21]](p1) :: (load (s8) from unknown-address + 22, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD21]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD22]](p1) :: (load (s8) from unknown-address + 23, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[SHL16]], [[ZEXTLOAD17]]
- ; GFX10-MESA-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[OR16]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[SHL17]], [[OR15]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR11]](s32), [[OR14]](s32), [[OR17]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-HSA-LABEL: name: test_global_v2s96_align1
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -27955,74 +24499,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-HSA-LABEL: name: test_global_v2s96_align2
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_global_v2s96_align2
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD1]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD1]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[ZEXTLOAD1]]
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
- ; GFX10-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD2:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD3]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LOAD2]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[SHL2]], [[ZEXTLOAD2]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD3:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD6]](p1) :: (load (s16) from unknown-address + 14, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LOAD3]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[SHL3]], [[ZEXTLOAD3]]
- ; GFX10-MESA-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C2]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s16) from unknown-address + 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s16) from unknown-address + 18, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
- ; GFX10-MESA-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD5]], [[C3]](s64)
- ; GFX10-MESA-NEXT: [[ZEXTLOAD5:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD9]](p1) :: (load (s16) from unknown-address + 20, addrspace 1)
- ; GFX10-MESA-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD9]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD10]](p1) :: (load (s16) from unknown-address + 22, addrspace 1)
- ; GFX10-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[SHL5]], [[ZEXTLOAD5]]
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR3]](s32), [[OR4]](s32), [[OR5]](s32)
- ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-HSA-LABEL: name: test_global_v2s96_align2
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -28271,36 +24747,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-HSA-LABEL: name: test_global_v2s96_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_global_v2s96_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-HSA-LABEL: name: test_global_v2s96_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -28471,36 +24917,6 @@ body: |
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
; GFX9-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
;
- ; GFX10-HSA-LABEL: name: test_global_v2s96_align16
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-HSA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-HSA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-HSA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
- ; GFX10-MESA-LABEL: name: test_global_v2s96_align16
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
- ; GFX10-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1)
- ; GFX10-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>)
- ; GFX10-MESA-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
- ; GFX10-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
- ; GFX10-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
- ; GFX10-MESA-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
- ;
; GFX11-HSA-LABEL: name: test_global_v2s96_align16
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -29096,244 +25512,6 @@ body: |
; GFX9-MESA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v32s1_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
- ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX10-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
- ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
- ; GFX10-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
- ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
- ; GFX10-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
- ; GFX10-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
- ; GFX10-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
- ; GFX10-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-HSA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
- ; GFX10-HSA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
- ; GFX10-HSA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
- ; GFX10-HSA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
- ; GFX10-HSA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
- ; GFX10-HSA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
- ; GFX10-HSA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
- ; GFX10-HSA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
- ; GFX10-HSA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
- ; GFX10-HSA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
- ; GFX10-HSA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
- ; GFX10-HSA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
- ; GFX10-HSA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
- ; GFX10-HSA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
- ; GFX10-HSA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
- ; GFX10-HSA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
- ; GFX10-HSA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
- ; GFX10-HSA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
- ; GFX10-HSA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
- ; GFX10-HSA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
- ; GFX10-HSA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
- ; GFX10-HSA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
- ; GFX10-HSA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
- ; GFX10-HSA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
- ; GFX10-HSA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
- ; GFX10-HSA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
- ; GFX10-HSA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
- ; GFX10-HSA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
- ; GFX10-HSA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
- ; GFX10-HSA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
- ; GFX10-HSA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-HSA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
- ; GFX10-HSA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
- ; GFX10-HSA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
- ; GFX10-HSA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
- ; GFX10-HSA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
- ; GFX10-HSA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
- ; GFX10-HSA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
- ; GFX10-HSA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
- ; GFX10-HSA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
- ; GFX10-HSA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
- ; GFX10-HSA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
- ; GFX10-HSA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
- ; GFX10-HSA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
- ; GFX10-HSA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
- ; GFX10-HSA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
- ; GFX10-HSA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v32s1_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
- ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX10-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
- ; GFX10-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
- ; GFX10-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
- ; GFX10-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
- ; GFX10-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C7]](s32)
- ; GFX10-MESA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 9
- ; GFX10-MESA-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C8]](s32)
- ; GFX10-MESA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
- ; GFX10-MESA-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C9]](s32)
- ; GFX10-MESA-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
- ; GFX10-MESA-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C10]](s32)
- ; GFX10-MESA-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
- ; GFX10-MESA-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C11]](s32)
- ; GFX10-MESA-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
- ; GFX10-MESA-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C12]](s32)
- ; GFX10-MESA-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
- ; GFX10-MESA-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C13]](s32)
- ; GFX10-MESA-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
- ; GFX10-MESA-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C14]](s32)
- ; GFX10-MESA-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C15]](s32)
- ; GFX10-MESA-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 17
- ; GFX10-MESA-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C16]](s32)
- ; GFX10-MESA-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 18
- ; GFX10-MESA-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C17]](s32)
- ; GFX10-MESA-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 19
- ; GFX10-MESA-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C18]](s32)
- ; GFX10-MESA-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
- ; GFX10-MESA-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C19]](s32)
- ; GFX10-MESA-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 21
- ; GFX10-MESA-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C20]](s32)
- ; GFX10-MESA-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 22
- ; GFX10-MESA-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C21]](s32)
- ; GFX10-MESA-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
- ; GFX10-MESA-NEXT: [[LSHR22:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C22]](s32)
- ; GFX10-MESA-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-MESA-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C23]](s32)
- ; GFX10-MESA-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 25
- ; GFX10-MESA-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C24]](s32)
- ; GFX10-MESA-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 26
- ; GFX10-MESA-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C25]](s32)
- ; GFX10-MESA-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 27
- ; GFX10-MESA-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C26]](s32)
- ; GFX10-MESA-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
- ; GFX10-MESA-NEXT: [[LSHR27:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C27]](s32)
- ; GFX10-MESA-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 29
- ; GFX10-MESA-NEXT: [[LSHR28:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C28]](s32)
- ; GFX10-MESA-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 30
- ; GFX10-MESA-NEXT: [[LSHR29:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C29]](s32)
- ; GFX10-MESA-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 31
- ; GFX10-MESA-NEXT: [[LSHR30:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C30]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR14]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR17]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR19]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR22]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC24:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR23]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC25:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR25]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR26]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR27]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR28]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC30:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR29]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC31:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR30]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC10]](s16), [[TRUNC11]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC12]](s16), [[TRUNC13]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC14]](s16), [[TRUNC15]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC16]](s16), [[TRUNC17]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR9:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC18]](s16), [[TRUNC19]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR10:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC20]](s16), [[TRUNC21]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC22]](s16), [[TRUNC23]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC24]](s16), [[TRUNC25]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC26]](s16), [[TRUNC27]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC28]](s16), [[TRUNC29]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR15:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC30]](s16), [[TRUNC31]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<32 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>), [[BUILD_VECTOR6]](<2 x s16>), [[BUILD_VECTOR7]](<2 x s16>), [[BUILD_VECTOR8]](<2 x s16>), [[BUILD_VECTOR9]](<2 x s16>), [[BUILD_VECTOR10]](<2 x s16>), [[BUILD_VECTOR11]](<2 x s16>), [[BUILD_VECTOR12]](<2 x s16>), [[BUILD_VECTOR13]](<2 x s16>), [[BUILD_VECTOR14]](<2 x s16>), [[BUILD_VECTOR15]](<2 x s16>)
- ; GFX10-MESA-NEXT: [[TRUNC32:%[0-9]+]]:_(<32 x s1>) = G_TRUNC [[CONCAT_VECTORS]](<32 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC32]](<32 x s1>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v32s1_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -29982,76 +26160,6 @@ body: |
; GFX9-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
;
- ; GFX10-HSA-LABEL: name: test_load_global_v8s4_align4
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX10-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-HSA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
- ; GFX10-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-HSA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-HSA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
- ; GFX10-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
- ; GFX10-HSA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
- ; GFX10-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-HSA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
- ; GFX10-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
- ; GFX10-HSA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
- ; GFX10-HSA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-HSA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
- ; GFX10-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
- ; GFX10-HSA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_v8s4_align4
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
- ; GFX10-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
- ; GFX10-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX10-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX10-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX10-MESA-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
- ; GFX10-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C2]](s32)
- ; GFX10-MESA-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-MESA-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C3]](s32)
- ; GFX10-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
- ; GFX10-MESA-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C4]](s32)
- ; GFX10-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
- ; GFX10-MESA-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C5]](s32)
- ; GFX10-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
- ; GFX10-MESA-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C6]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
- ; GFX10-MESA-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
- ; GFX10-MESA-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16)
- ; GFX10-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
- ; GFX10-MESA-NEXT: [[TRUNC8:%[0-9]+]]:_(<8 x s4>) = G_TRUNC [[CONCAT_VECTORS]](<8 x s16>)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[TRUNC8]](<8 x s4>)
- ;
; GFX11-HSA-LABEL: name: test_load_global_v8s4_align4
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
@@ -30245,20 +26353,6 @@ body: |
; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
; GFX9-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; GFX10-HSA-LABEL: name: test_load_global_s32_align536870912
- ; GFX10-HSA: liveins: $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: {{ $}}
- ; GFX10-HSA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
- ; GFX10-HSA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; GFX10-MESA-LABEL: name: test_load_global_s32_align536870912
- ; GFX10-MESA: liveins: $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: {{ $}}
- ; GFX10-MESA-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 536870912, addrspace 1)
- ; GFX10-MESA-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
; GFX11-HSA-LABEL: name: test_load_global_s32_align536870912
; GFX11-HSA: liveins: $vgpr0_vgpr1
; GFX11-HSA-NEXT: {{ $}}
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir
index f369d588b509b..623c223c8a165 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir
@@ -8643,125 +8643,233 @@ body: |
; GFX10-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX10-UNALIGNED-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX11-LABEL: name: test_load_local_v3s8_align4
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX11-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_v3s8_align4
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
- ; GFX11-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-UNALIGNED-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX11-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-UNALIGNED-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX11-UNALIGNED-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX11-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX11-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX11-UNALIGNED-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-UNALIGNED-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX11-UNALIGNED-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX11-UNALIGNED-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX11-UNALIGNED-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX11-UNALIGNED-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-UNALIGNED-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX12-LABEL: name: test_load_local_v3s8_align4
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX12-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s8_align4
- ; GFX12-UNALIGNED: liveins: $vgpr0
- ; GFX12-UNALIGNED-NEXT: {{ $}}
- ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
- ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX12-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-UNALIGNED-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX12-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-UNALIGNED-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX12-UNALIGNED-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX12-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX12-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-UNALIGNED-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX12-UNALIGNED-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-UNALIGNED-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX12-UNALIGNED-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX12-UNALIGNED-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX12-UNALIGNED-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX12-UNALIGNED-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-UNALIGNED-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s8_align4
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 1)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), addrspace 1, align 4)
%2:_(s24) = G_BITCAST %1
@@ -9091,155 +9199,293 @@ body: |
; GFX10-UNALIGNED-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
; GFX10-UNALIGNED-NEXT: $vgpr0 = COPY [[OR3]](s32)
;
- ; GFX11-LABEL: name: test_load_local_v3s8_align1
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
- ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; GFX11-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; GFX11-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_v3s8_align1
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-UNALIGNED-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX11-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX11-UNALIGNED-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX11-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-UNALIGNED-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX11-UNALIGNED-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-UNALIGNED-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX11-UNALIGNED-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX11-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX11-UNALIGNED-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-UNALIGNED-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX11-UNALIGNED-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX11-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX11-UNALIGNED-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-UNALIGNED-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX11-UNALIGNED-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-UNALIGNED-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX11-UNALIGNED-NEXT: $vgpr0 = COPY [[OR3]](s32)
- ;
- ; GFX12-LABEL: name: test_load_local_v3s8_align1
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
- ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; GFX12-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; GFX12-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
- ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s8_align1
- ; GFX12-UNALIGNED: liveins: $vgpr0
- ; GFX12-UNALIGNED-NEXT: {{ $}}
- ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX12-UNALIGNED-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
- ; GFX12-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX12-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
- ; GFX12-UNALIGNED-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-UNALIGNED-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX12-UNALIGNED-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX12-UNALIGNED-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX12-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-UNALIGNED-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX12-UNALIGNED-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-UNALIGNED-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX12-UNALIGNED-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-UNALIGNED-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX12-UNALIGNED-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-UNALIGNED-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX12-UNALIGNED-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX12-UNALIGNED-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-UNALIGNED-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX12-UNALIGNED-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-UNALIGNED-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX12-UNALIGNED-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX12-UNALIGNED-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX12-UNALIGNED-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-UNALIGNED-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX12-UNALIGNED-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-UNALIGNED-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX12-UNALIGNED-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX11-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX11-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX11-FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX12-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX12-TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX12-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s8), addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 1, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX12-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX12-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX12-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX12-FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s8_align1
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 3)
%2:_(s24) = G_BITCAST %1
@@ -10646,49 +10892,101 @@ body: |
; GFX10-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
; GFX10-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-LABEL: name: test_load_local_v3s16_align8
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
- ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX11-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-UNALIGNED-LABEL: name: test_load_local_v3s16_align8
- ; GFX11-UNALIGNED: liveins: $vgpr0
- ; GFX11-UNALIGNED-NEXT: {{ $}}
- ; GFX11-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX11-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
- ; GFX11-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX11-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX11-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-FAKE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX11-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX12-LABEL: name: test_load_local_v3s16_align8
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
- ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX11-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX12-UNALIGNED-LABEL: name: test_load_local_v3s16_align8
- ; GFX12-UNALIGNED: liveins: $vgpr0
- ; GFX12-UNALIGNED-NEXT: {{ $}}
- ; GFX12-UNALIGNED-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
- ; GFX12-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
- ; GFX12-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX12-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX12-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX11-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-UNALIGNED-TRUE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX12-UNALIGNED-TRUE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-UNALIGNED-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-UNALIGNED-FAKE16-LABEL: name: test_load_local_v3s16_align8
+ ; GFX12-UNALIGNED-FAKE16: liveins: $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: {{ $}}
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-UNALIGNED-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p3) = COPY $vgpr0
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 3)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir
index 52dc8701e86f2..be5ef98a6f395 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir
@@ -7928,65 +7928,119 @@ body: |
; GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; GFX11-LABEL: name: test_load_private_v3s8_align4
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX11-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; GFX12-LABEL: name: test_load_private_v3s8_align4
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; GFX12-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_v3s8_align4
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_v3s8_align4
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX11-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX11-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX11-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_v3s8_align4
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_v3s8_align4
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_v3s8_align4
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -8048,65 +8102,119 @@ body: |
; UNALIGNED_GFX10-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; UNALIGNED_GFX10-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_v3s8_align4
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; UNALIGNED_GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; UNALIGNED_GFX11-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX11-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; UNALIGNED_GFX11-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; UNALIGNED_GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; UNALIGNED_GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; UNALIGNED_GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; UNALIGNED_GFX11-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; UNALIGNED_GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; UNALIGNED_GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; UNALIGNED_GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; UNALIGNED_GFX11-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_v3s8_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; UNALIGNED_GFX12-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX12-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; UNALIGNED_GFX12-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; UNALIGNED_GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; UNALIGNED_GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; UNALIGNED_GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; UNALIGNED_GFX12-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; UNALIGNED_GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; UNALIGNED_GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; UNALIGNED_GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_v3s8_align4
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_v3s8_align4
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_v3s8_align4
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_v3s8_align4
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
%0:_(p5) = COPY $vgpr0
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), addrspace 5, align 4)
%2:_(s24) = G_BITCAST %1
@@ -8314,75 +8422,139 @@ body: |
; GFX10-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
; GFX10-NEXT: $vgpr0 = COPY [[OR3]](s32)
;
- ; GFX11-LABEL: name: test_load_private_v3s8_align1
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
- ; GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX11-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX11-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX11-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX11-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX11-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX11-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX11-NEXT: $vgpr0 = COPY [[OR3]](s32)
- ;
- ; GFX12-LABEL: name: test_load_private_v3s8_align1
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
- ; GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
- ; GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
- ; GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
- ; GFX12-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
- ; GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
- ; GFX12-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
- ; GFX12-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
- ; GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
- ; GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
- ; GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
- ; GFX12-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
- ; GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; GFX12-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
- ; GFX12-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_v3s8_align1
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX11-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX11-FAKE16-LABEL: name: test_load_private_v3s8_align1
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX11-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX11-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX11-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX11-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX11-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX11-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX11-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX11-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX11-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX11-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX11-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX11-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX11-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX11-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX11-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX11-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX11-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX11-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_v3s8_align1
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR1]](s16), [[OR2]](s16)
+ ; GFX12-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_v3s8_align1
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s16), align 1, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32)
+ ; GFX12-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR]](s32)
+ ; GFX12-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C4]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL1]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C3]]
+ ; GFX12-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C4]](s16)
+ ; GFX12-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL2]]
+ ; GFX12-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
+ ; GFX12-FAKE16-NEXT: $vgpr0 = COPY [[OR3]](s32)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_v3s8_align1
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -8464,85 +8636,159 @@ body: |
; UNALIGNED_GFX10-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; UNALIGNED_GFX10-NEXT: $vgpr0 = COPY [[OR4]](s32)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_v3s8_align1
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s8), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX11-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 1, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX11-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX11-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX11-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX11-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; UNALIGNED_GFX11-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; UNALIGNED_GFX11-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX11-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX11-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; UNALIGNED_GFX11-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX11-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; UNALIGNED_GFX11-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX11-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; UNALIGNED_GFX11-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; UNALIGNED_GFX11-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX11-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; UNALIGNED_GFX11-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX11-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; UNALIGNED_GFX11-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; UNALIGNED_GFX11-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; UNALIGNED_GFX11-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; UNALIGNED_GFX11-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; UNALIGNED_GFX11-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; UNALIGNED_GFX11-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; UNALIGNED_GFX11-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_v3s8_align1
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s8), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
- ; UNALIGNED_GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 1, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX12-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX12-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; UNALIGNED_GFX12-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; UNALIGNED_GFX12-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX12-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX12-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX12-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; UNALIGNED_GFX12-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX12-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; UNALIGNED_GFX12-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; UNALIGNED_GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; UNALIGNED_GFX12-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX12-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; UNALIGNED_GFX12-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; UNALIGNED_GFX12-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; UNALIGNED_GFX12-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; UNALIGNED_GFX12-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; UNALIGNED_GFX12-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; UNALIGNED_GFX12-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_v3s8_align1
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s8), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 1, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_v3s8_align1
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s8), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 1, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_v3s8_align1
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s8), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 1, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_v3s8_align1
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p5) :: (load (s8), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p5) :: (load (s8) from unknown-address + 1, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s8) from unknown-address + 2, addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
%0:_(p5) = COPY $vgpr0
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 5)
%2:_(s24) = G_BITCAST %1
@@ -9879,27 +10125,53 @@ body: |
; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX11-LABEL: name: test_load_private_v3s16_align8
- ; GFX11: liveins: $vgpr0
- ; GFX11-NEXT: {{ $}}
- ; GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
- ; GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-TRUE16-LABEL: name: test_load_private_v3s16_align8
+ ; GFX11-TRUE16: liveins: $vgpr0
+ ; GFX11-TRUE16-NEXT: {{ $}}
+ ; GFX11-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; GFX11-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX11-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; GFX12-LABEL: name: test_load_private_v3s16_align8
- ; GFX12: liveins: $vgpr0
- ; GFX12-NEXT: {{ $}}
- ; GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
- ; GFX12-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; GFX11-FAKE16-LABEL: name: test_load_private_v3s16_align8
+ ; GFX11-FAKE16: liveins: $vgpr0
+ ; GFX11-FAKE16-NEXT: {{ $}}
+ ; GFX11-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX11-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; GFX11-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX11-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX11-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX11-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX11-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-TRUE16-LABEL: name: test_load_private_v3s16_align8
+ ; GFX12-TRUE16: liveins: $vgpr0
+ ; GFX12-TRUE16-NEXT: {{ $}}
+ ; GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; GFX12-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-FAKE16-LABEL: name: test_load_private_v3s16_align8
+ ; GFX12-FAKE16: liveins: $vgpr0
+ ; GFX12-FAKE16-NEXT: {{ $}}
+ ; GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; GFX12-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
; UNALIGNED_GFX9-LABEL: name: test_load_private_v3s16_align8
; UNALIGNED_GFX9: liveins: $vgpr0
@@ -9943,27 +10215,53 @@ body: |
; UNALIGNED_GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>)
; UNALIGNED_GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; UNALIGNED_GFX11-LABEL: name: test_load_private_v3s16_align8
- ; UNALIGNED_GFX11: liveins: $vgpr0
- ; UNALIGNED_GFX11-NEXT: {{ $}}
- ; UNALIGNED_GFX11-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX11-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
- ; UNALIGNED_GFX11-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; UNALIGNED_GFX11-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX11-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; UNALIGNED_GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-LABEL: name: test_load_private_v3s16_align8
+ ; UNALIGNED_GFX11_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX11_TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
;
- ; UNALIGNED_GFX12-LABEL: name: test_load_private_v3s16_align8
- ; UNALIGNED_GFX12: liveins: $vgpr0
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
- ; UNALIGNED_GFX12-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; UNALIGNED_GFX12-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX12-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX12-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
- ; UNALIGNED_GFX12-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ; UNALIGNED_GFX11_FAKE16-LABEL: name: test_load_private_v3s16_align8
+ ; UNALIGNED_GFX11_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; UNALIGNED_GFX11_FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX12_TRUE16-LABEL: name: test_load_private_v3s16_align8
+ ; UNALIGNED_GFX12_TRUE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV4]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX12_TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX12_FAKE16-LABEL: name: test_load_private_v3s16_align8
+ ; UNALIGNED_GFX12_FAKE16: liveins: $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p5) :: (load (<4 x s16>), addrspace 5)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[BUILD_VECTOR]](<2 x s16>)
+ ; UNALIGNED_GFX12_FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
%0:_(p5) = COPY $vgpr0
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 5)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
index 423e60b5dc5b5..45cfd6f315f1a 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
@@ -3,7 +3,7 @@
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=CI %s
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX10 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -mattr=+real-true16 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
@@ -55,16 +55,6 @@ body: |
; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[AND]], [[C]]
; GFX9-NEXT: G_STORE [[AND1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s1_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
- ; GFX10-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[AND]], [[C]]
- ; GFX10-NEXT: G_STORE [[AND1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s1_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -132,15 +122,6 @@ body: |
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
; GFX9-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s7_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 127
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
- ; GFX10-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s7_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -198,13 +179,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s8_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s8), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s8_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -272,13 +246,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s16_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
- ;
; GFX11-TRUE16-LABEL: name: test_store_global_s16_align1
; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-TRUE16-NEXT: {{ $}}
@@ -348,13 +315,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s16_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
- ;
; GFX11-TRUE16-LABEL: name: test_store_global_s16_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-TRUE16-NEXT: {{ $}}
@@ -424,13 +384,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
- ;
; GFX11-TRUE16-LABEL: name: test_store_global_s16_align4
; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-TRUE16-NEXT: {{ $}}
@@ -524,19 +477,6 @@ body: |
; GFX9-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
; GFX9-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s24_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1)
- ; GFX10-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s24_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -626,19 +566,6 @@ body: |
; GFX9-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
; GFX9-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s24_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
- ; GFX10-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s24_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -742,19 +669,6 @@ body: |
; GFX9-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
; GFX9-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s24_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), align 1, addrspace 1)
- ; GFX10-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s8) into unknown-address + 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s24_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -828,15 +742,6 @@ body: |
; GFX9-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
; GFX9-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s25_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 33554431
- ; GFX10-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
- ; GFX10-NEXT: G_STORE [[AND]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s25_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -953,13 +858,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s32_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -1024,13 +922,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s32_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -1083,13 +974,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](s32), [[COPY]](p1) :: (store (s32), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s32_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -1179,13 +1063,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p3_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p3_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -1252,13 +1129,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p3_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p3_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -1311,13 +1181,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
; GFX9-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p3_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY $vgpr2
- ; GFX10-NEXT: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store (p3), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p3_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2
; GFX11-NEXT: {{ $}}
@@ -1447,21 +1310,6 @@ body: |
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
; GFX9-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s48_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
- ; GFX10-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 1, addrspace 1)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
- ; GFX10-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, align 1, addrspace 1)
- ;
; GFX11-TRUE16-LABEL: name: test_store_global_s48_align1
; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-TRUE16-NEXT: {{ $}}
@@ -1605,21 +1453,6 @@ body: |
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
; GFX9-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s48_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[COPY1]](s64)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
- ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY2]], [[C]](s32)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY2]](s64)
- ; GFX10-NEXT: G_STORE [[TRUNC]](s32), [[COPY]](p1) :: (store (s32), align 2, addrspace 1)
- ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s32) = G_TRUNC [[LSHR]](s64)
- ; GFX10-NEXT: G_STORE [[TRUNC1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
- ;
; GFX11-TRUE16-LABEL: name: test_store_global_s48_align2
; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-TRUE16-NEXT: {{ $}}
@@ -1822,13 +1655,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s64_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s64_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -1917,13 +1743,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s64_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s64_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -1976,13 +1795,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s64_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s64_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2035,13 +1847,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s64_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s64_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2094,13 +1899,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s64_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](s64), [[COPY]](p1) :: (store (s64), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s64_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2233,13 +2031,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p0_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p0_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2330,13 +2121,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p0_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p0_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2389,13 +2173,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p0_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p0_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2448,13 +2225,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p0_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p0_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2507,13 +2277,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p0_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p0) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p0), [[COPY]](p1) :: (store (p0), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p0_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2646,13 +2409,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p999_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p999_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2743,13 +2499,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p999_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p999_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2802,13 +2551,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p999_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p999_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2861,13 +2603,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p999_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p999_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -2920,13 +2655,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_p999_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(p999) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](p999), [[COPY]](p1) :: (store (p999), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_p999_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3049,13 +2777,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s32_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3136,13 +2857,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s32_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3195,13 +2909,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s32_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3254,13 +2961,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s32_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3313,13 +3013,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s32_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3446,14 +3139,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p3_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p3_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3540,14 +3225,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p3_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p3_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3606,14 +3283,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p3_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p3_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3672,14 +3341,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p3_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p3_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3738,14 +3399,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
; GFX9-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p3_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p3>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY1]](<2 x p3>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p3_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3874,13 +3527,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s16_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s16_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -3963,13 +3609,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s16_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s16_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -4022,13 +3661,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s16_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -4081,13 +3713,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s16_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s16_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -4140,13 +3765,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s16_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s16>), [[COPY]](p1) :: (store (<4 x s16>), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s16_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX11-NEXT: {{ $}}
@@ -4302,13 +3920,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v3s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v3s32_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -4403,13 +4014,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v3s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v3s32_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -4467,13 +4071,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v3s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v3s32_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -4531,13 +4128,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v3s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v3s32_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -4595,13 +4185,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
; GFX9-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v3s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v3s32_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -4790,13 +4373,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s32_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -4905,13 +4481,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s32_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -4964,13 +4533,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s32_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5023,13 +4585,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s32_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5082,13 +4637,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v4s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v4s32_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5295,13 +4843,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s64_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s64_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5428,13 +4969,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s64_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s64_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5487,13 +5021,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s64_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s64_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5546,13 +5073,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s64_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s64_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5605,13 +5125,6 @@ body: |
; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
; GFX9-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s64_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: G_STORE [[COPY1]](<2 x s64>), [[COPY]](p1) :: (store (<2 x s64>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s64_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5804,14 +5317,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s16_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s16_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5926,14 +5431,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s16_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s16_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -5992,14 +5489,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s16_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6058,14 +5547,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s16_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s16_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6124,14 +5605,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s16_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s16_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6326,14 +5799,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p0_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p0_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6448,14 +5913,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p0_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p0_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6514,14 +5971,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p0_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p0_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6580,14 +6029,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p0_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p0_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6646,14 +6087,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2p0_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2p0_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -6815,14 +6248,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s96_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s96_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -6923,14 +6348,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s96_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s96_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -6994,14 +6411,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s96_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s96_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -7065,14 +6474,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s96_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s96_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -7136,14 +6537,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
; GFX9-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s96_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s96_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
; GFX11-NEXT: {{ $}}
@@ -7338,14 +6731,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s128_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s128_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -7460,14 +6845,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s128_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s128_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -7526,14 +6903,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s128_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s128_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -7592,14 +6961,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s128_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s128_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -7658,14 +7019,6 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s128_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s128_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5
; GFX11-NEXT: {{ $}}
@@ -7899,18 +7252,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5s32_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8053,18 +7394,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5s32_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8147,18 +7476,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5s32_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8241,18 +7558,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5s32_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8335,18 +7640,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5s32_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8591,19 +7884,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5p3_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5p3_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8752,19 +8032,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5p3_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5p3_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8853,19 +8120,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5p3_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5p3_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -8954,19 +8208,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5p3_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5p3_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -9055,19 +8296,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v5p3_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v5p3_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -9156,19 +8384,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v10s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v10s16_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -9324,29 +8539,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
; GFX9-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v11s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>), [[BUILD_VECTOR2]](<2 x s16>), [[BUILD_VECTOR3]](<2 x s16>)
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
- ; GFX10-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; GFX10-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
- ; GFX10-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; GFX10-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
- ; GFX10-NEXT: G_STORE [[DEF1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
- ;
; GFX11-TRUE16-LABEL: name: test_store_global_v11s16_align4
; GFX11-TRUE16: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-TRUE16-NEXT: {{ $}}
@@ -9499,19 +8691,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV2]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v12s16_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>), [[UV2:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
- ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[UV]](<2 x s32>), [[UV1]](<2 x s32>)
- ; GFX10-NEXT: G_STORE [[CONCAT_VECTORS]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV2]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v12s16_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -9759,19 +8938,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s160_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s160_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -9920,19 +9086,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s160_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s160_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -10021,19 +9174,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s160_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s160_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -10122,19 +9262,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s160_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s160_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -10223,19 +9350,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s160_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s160_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
; GFX11-NEXT: {{ $}}
@@ -10572,17 +9686,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s32_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -10759,17 +9862,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s32_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -10846,17 +9938,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s32_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -10933,17 +10014,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s32_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -11020,17 +10090,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s32_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -11111,18 +10170,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v2s128_align32
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s128>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](<2 x s128>)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v2s128_align32
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -11461,18 +10508,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s256_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s256_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -11655,18 +10690,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s256_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s256_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -11749,18 +10772,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s256_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s256_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -11843,18 +10854,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s256_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s256_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -11937,18 +10936,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s256_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s256_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -12031,18 +11018,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_s256_align32
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s256) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s32>) = G_BITCAST [[COPY1]](s256)
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_s256_align32
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -12121,17 +11096,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
; GFX9-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v8s32_align32
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[COPY1]](<8 x s32>)
- ; GFX10-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[UV1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v8s32_align32
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9
; GFX11-NEXT: {{ $}}
@@ -12521,26 +11485,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v9s32_align1
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
- ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
- ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
- ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v9s32_align1
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
; GFX11-NEXT: {{ $}}
@@ -12776,26 +11720,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v9s32_align2
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
- ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
- ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
- ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v9s32_align2
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
; GFX11-NEXT: {{ $}}
@@ -12929,26 +11853,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v9s32_align4
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
- ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
- ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
- ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v9s32_align4
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
; GFX11-NEXT: {{ $}}
@@ -13082,26 +11986,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v9s32_align8
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
- ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
- ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
- ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v9s32_align8
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
; GFX11-NEXT: {{ $}}
@@ -13235,26 +12119,6 @@ body: |
; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
;
- ; GFX10-LABEL: name: test_store_global_v9s32_align16
- ; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
- ; GFX10-NEXT: {{ $}}
- ; GFX10-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
- ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
- ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
- ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
- ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
- ; GFX10-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
- ; GFX10-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
- ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
- ; GFX10-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
- ; GFX10-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
- ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; GFX10-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
- ; GFX10-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
- ; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; GFX10-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
- ;
; GFX11-LABEL: name: test_store_global_v9s32_align16
; GFX11: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
; GFX11-NEXT: {{ $}}
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll
index 49cc38a0af276..a30e3115cc463 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/uaddsat.ll
@@ -340,11 +340,8 @@ define i16 @v_uaddsat_v2i8(i16 %lhs.arg, i16 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, v1 clamp
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v0 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
-; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
-; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
-; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_uaddsat_v2i8:
@@ -445,33 +442,86 @@ define amdgpu_ps i16 @s_uaddsat_v2i8(i16 inreg %lhs.arg, i16 inreg %rhs.arg) {
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
-; GFX10PLUS-LABEL: s_uaddsat_v2i8:
-; GFX10PLUS: ; %bb.0:
-; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 8
-; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 8
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 16
-; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 16
-; GFX10PLUS-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX10PLUS-NEXT: s_lshl_b32 s2, s2, 8
-; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX10PLUS-NEXT: s_lshl_b32 s3, s3, 8
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10PLUS-NEXT: v_pk_add_u16 v0, s0, s1 clamp
-; GFX10PLUS-NEXT: v_readfirstlane_b32 s0, v0
-; GFX10PLUS-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 16
-; GFX10PLUS-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 8
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX10PLUS-NEXT: s_lshr_b32 s1, s0, 16
-; GFX10PLUS-NEXT: s_and_b32 s0, s0, 0xff
-; GFX10PLUS-NEXT: s_and_b32 s1, s1, 0xff
-; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 8
-; GFX10PLUS-NEXT: s_or_b32 s0, s0, s1
-; GFX10PLUS-NEXT: ; return to shader part epilog
+; GFX10-LABEL: s_uaddsat_v2i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_lshr_b32 s2, s0, 8
+; GFX10-NEXT: s_lshr_b32 s3, s1, 8
+; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10-NEXT: s_lshr_b32 s2, s0, 16
+; GFX10-NEXT: s_lshr_b32 s3, s1, 16
+; GFX10-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX10-NEXT: s_lshl_b32 s2, s2, 8
+; GFX10-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX10-NEXT: s_lshl_b32 s3, s3, 8
+; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10-NEXT: v_pk_add_u16 v0, s0, s1 clamp
+; GFX10-NEXT: v_readfirstlane_b32 s0, v0
+; GFX10-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX10-NEXT: s_lshr_b32 s0, s0, 16
+; GFX10-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX10-NEXT: s_lshr_b32 s0, s0, 8
+; GFX10-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX10-NEXT: s_lshr_b32 s1, s0, 16
+; GFX10-NEXT: s_and_b32 s0, s0, 0xff
+; GFX10-NEXT: s_and_b32 s1, s1, 0xff
+; GFX10-NEXT: s_lshl_b32 s1, s1, 8
+; GFX10-NEXT: s_or_b32 s0, s0, s1
+; GFX10-NEXT: ; return to shader part epilog
+;
+; GFX11-TRUE16-LABEL: s_uaddsat_v2i8:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, s0, s1 clamp
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-TRUE16-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s0, 8
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX11-FAKE16-LABEL: s_uaddsat_v2i8:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v0, s0, s1 clamp
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-FAKE16-NEXT: ; return to shader part epilog
%lhs = bitcast i16 %lhs.arg to <2 x i8>
%rhs = bitcast i16 %rhs.arg to <2 x i8>
%result = call <2 x i8> @llvm.uadd.sat.v2i8(<2 x i8> %lhs, <2 x i8> %rhs)
@@ -629,19 +679,21 @@ define i32 @v_uaddsat_v4i8(i32 %lhs.arg, i32 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v0, 8, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v2, 8, v3 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v3, 8, v6 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, v1 clamp
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v2, v3 clamp
+; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v6 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v0, 8, v0 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v2, v1 clamp
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 8
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v1
-; GFX11-TRUE16-NEXT: v_bfe_u32 v1, v1, 16, 8
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v1.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 24, v1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v3
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-TRUE16-NEXT: v_and_or_b32 v0, 0xff, v0, v2
-; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v3, v1
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 24, v3
+; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v2
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_uaddsat_v4i8:
@@ -876,57 +928,107 @@ define amdgpu_ps i32 @s_uaddsat_v4i8(i32 inreg %lhs.arg, i32 inreg %rhs.arg) {
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
-; GFX11-LABEL: s_uaddsat_v4i8:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-NEXT: s_lshr_b32 s3, s0, 24
-; GFX11-NEXT: s_lshr_b32 s4, s1, 8
-; GFX11-NEXT: s_lshr_b32 s5, s1, 24
-; GFX11-NEXT: s_pack_ll_b32_b16 s2, s0, s2
-; GFX11-NEXT: s_pack_hl_b32_b16 s0, s0, s3
-; GFX11-NEXT: s_pack_ll_b32_b16 s3, s1, s4
-; GFX11-NEXT: s_lshr_b32 s4, s2, 16
-; GFX11-NEXT: s_pack_hl_b32_b16 s1, s1, s5
-; GFX11-NEXT: s_lshr_b32 s5, s3, 16
-; GFX11-NEXT: s_lshl_b32 s2, s2, 0x80008
-; GFX11-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-NEXT: s_lshl_b32 s3, s3, 0x80008
-; GFX11-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-NEXT: s_pack_ll_b32_b16 s2, s2, s4
-; GFX11-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-NEXT: s_pack_ll_b32_b16 s3, s3, s5
-; GFX11-NEXT: s_lshr_b32 s5, s1, 16
-; GFX11-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-NEXT: v_pk_add_u16 v0, s2, s3 clamp
-; GFX11-NEXT: s_pack_ll_b32_b16 s0, s0, s4
-; GFX11-NEXT: s_pack_ll_b32_b16 s1, s1, s5
-; GFX11-NEXT: v_pk_add_u16 v1, s0, s1 clamp
-; GFX11-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-NEXT: v_readfirstlane_b32 s1, v1
-; GFX11-NEXT: s_and_b32 s2, s0, 0xffff
-; GFX11-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-NEXT: s_and_b32 s3, s1, 0xffff
-; GFX11-NEXT: s_lshr_b32 s1, s1, 16
-; GFX11-NEXT: s_lshr_b32 s3, s3, 0x80008
-; GFX11-NEXT: s_lshr_b32 s1, s1, 8
-; GFX11-NEXT: s_pack_ll_b32_b16 s0, s2, s0
-; GFX11-NEXT: s_pack_ll_b32_b16 s1, s3, s1
-; GFX11-NEXT: s_bfe_u32 s2, s0, 0x80010
-; GFX11-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-NEXT: s_and_b32 s3, s1, 0xff
-; GFX11-NEXT: s_or_b32 s0, s0, s2
-; GFX11-NEXT: s_lshl_b32 s2, s3, 16
-; GFX11-NEXT: s_bfe_u32 s1, s1, 0x80010
-; GFX11-NEXT: s_or_b32 s0, s0, s2
-; GFX11-NEXT: s_lshl_b32 s1, s1, 24
-; GFX11-NEXT: s_or_b32 s0, s0, s1
-; GFX11-NEXT: ; return to shader part epilog
+; GFX11-TRUE16-LABEL: s_uaddsat_v4i8:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 24
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
+; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
+; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s2, 16
+; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
+; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s3, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
+; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
+; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, s2, s3 clamp
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, s0, s1 clamp
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s1, v1
+; GFX11-TRUE16-NEXT: s_and_b32 s2, s0, 0xffff
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
+; GFX11-TRUE16-NEXT: s_and_b32 s2, s1, 0xffff
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s2, s1
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s0, 8
+; GFX11-TRUE16-NEXT: s_and_b32 s1, s1, 0xff
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s1, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 24
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX11-FAKE16-LABEL: s_uaddsat_v4i8:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s1, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 24
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
+; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s2, 16
+; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v0, s2, s3 clamp
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v1, s0, s1 clamp
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s1, v1
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xffff
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s3, 0x80008
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s3, s1
+; GFX11-FAKE16-NEXT: s_bfe_u32 s2, s0, 0x80010
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s3, 16
+; GFX11-FAKE16-NEXT: s_bfe_u32 s1, s1, 0x80010
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 24
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-FAKE16-NEXT: ; return to shader part epilog
%lhs = bitcast i32 %lhs.arg to <4 x i8>
%rhs = bitcast i32 %rhs.arg to <4 x i8>
%result = call <4 x i8> @llvm.uadd.sat.v4i8(<4 x i8> %lhs, <4 x i8> %rhs)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll
index 8f248b4d2700c..1c8a26ce6126d 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/usubsat.ll
@@ -334,11 +334,8 @@ define i16 @v_usubsat_v2i8(i16 %lhs.arg, i16 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, v0, v1 clamp
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v0 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
-; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
-; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
-; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
+; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_usubsat_v2i8:
@@ -437,33 +434,86 @@ define amdgpu_ps i16 @s_usubsat_v2i8(i16 inreg %lhs.arg, i16 inreg %rhs.arg) {
; GFX9-NEXT: s_or_b32 s0, s0, s1
; GFX9-NEXT: ; return to shader part epilog
;
-; GFX10PLUS-LABEL: s_usubsat_v2i8:
-; GFX10PLUS: ; %bb.0:
-; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 8
-; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 8
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10PLUS-NEXT: s_lshr_b32 s2, s0, 16
-; GFX10PLUS-NEXT: s_lshr_b32 s3, s1, 16
-; GFX10PLUS-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX10PLUS-NEXT: s_lshl_b32 s2, s2, 8
-; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX10PLUS-NEXT: s_lshl_b32 s3, s3, 8
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s0, s2
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s1, s1, s3
-; GFX10PLUS-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
-; GFX10PLUS-NEXT: v_readfirstlane_b32 s0, v0
-; GFX10PLUS-NEXT: s_and_b32 s1, s0, 0xffff
-; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 16
-; GFX10PLUS-NEXT: s_lshr_b32 s1, s1, 0x80008
-; GFX10PLUS-NEXT: s_lshr_b32 s0, s0, 8
-; GFX10PLUS-NEXT: s_pack_ll_b32_b16 s0, s1, s0
-; GFX10PLUS-NEXT: s_lshr_b32 s1, s0, 16
-; GFX10PLUS-NEXT: s_and_b32 s0, s0, 0xff
-; GFX10PLUS-NEXT: s_and_b32 s1, s1, 0xff
-; GFX10PLUS-NEXT: s_lshl_b32 s1, s1, 8
-; GFX10PLUS-NEXT: s_or_b32 s0, s0, s1
-; GFX10PLUS-NEXT: ; return to shader part epilog
+; GFX10-LABEL: s_usubsat_v2i8:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_lshr_b32 s2, s0, 8
+; GFX10-NEXT: s_lshr_b32 s3, s1, 8
+; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10-NEXT: s_lshr_b32 s2, s0, 16
+; GFX10-NEXT: s_lshr_b32 s3, s1, 16
+; GFX10-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX10-NEXT: s_lshl_b32 s2, s2, 8
+; GFX10-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX10-NEXT: s_lshl_b32 s3, s3, 8
+; GFX10-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX10-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX10-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
+; GFX10-NEXT: v_readfirstlane_b32 s0, v0
+; GFX10-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX10-NEXT: s_lshr_b32 s0, s0, 16
+; GFX10-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX10-NEXT: s_lshr_b32 s0, s0, 8
+; GFX10-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX10-NEXT: s_lshr_b32 s1, s0, 16
+; GFX10-NEXT: s_and_b32 s0, s0, 0xff
+; GFX10-NEXT: s_and_b32 s1, s1, 0xff
+; GFX10-NEXT: s_lshl_b32 s1, s1, 8
+; GFX10-NEXT: s_or_b32 s0, s0, s1
+; GFX10-NEXT: ; return to shader part epilog
+;
+; GFX11-TRUE16-LABEL: s_usubsat_v2i8:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-TRUE16-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s0, 8
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX11-FAKE16-LABEL: s_usubsat_v2i8:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s3
+; GFX11-FAKE16-NEXT: v_pk_sub_u16 v0, s0, s1 clamp
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 0x80008
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s1, s0
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-FAKE16-NEXT: ; return to shader part epilog
%lhs = bitcast i16 %lhs.arg to <2 x i8>
%rhs = bitcast i16 %rhs.arg to <2 x i8>
%result = call <2 x i8> @llvm.usub.sat.v2i8(<2 x i8> %lhs, <2 x i8> %rhs)
@@ -617,19 +667,21 @@ define i32 @v_usubsat_v4i8(i32 %lhs.arg, i32 %rhs.arg) {
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v0, 8, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v2, 8, v3 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v3, 8, v6 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, v0, v1 clamp
-; GFX11-TRUE16-NEXT: v_pk_sub_u16 v1, v2, v3 clamp
+; GFX11-TRUE16-NEXT: v_pk_lshlrev_b16 v1, 8, v6 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v0, 8, v0 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_sub_u16 v1, v2, v1 clamp
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
; GFX11-TRUE16-NEXT: v_pk_lshrrev_b16 v1, 8, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 8
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v1
-; GFX11-TRUE16-NEXT: v_bfe_u32 v1, v1, 16, 8
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xff, v2
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v1.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 24, v1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xff, v3
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-TRUE16-NEXT: v_and_or_b32 v0, 0xff, v0, v2
-; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v3, v1
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 24, v3
+; GFX11-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v2
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_usubsat_v4i8:
@@ -860,57 +912,107 @@ define amdgpu_ps i32 @s_usubsat_v4i8(i32 inreg %lhs.arg, i32 inreg %rhs.arg) {
; GFX10-NEXT: s_or_b32 s0, s0, s1
; GFX10-NEXT: ; return to shader part epilog
;
-; GFX11-LABEL: s_usubsat_v4i8:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_lshr_b32 s2, s0, 8
-; GFX11-NEXT: s_lshr_b32 s3, s0, 24
-; GFX11-NEXT: s_lshr_b32 s4, s1, 8
-; GFX11-NEXT: s_lshr_b32 s5, s1, 24
-; GFX11-NEXT: s_pack_ll_b32_b16 s2, s0, s2
-; GFX11-NEXT: s_pack_hl_b32_b16 s0, s0, s3
-; GFX11-NEXT: s_pack_ll_b32_b16 s3, s1, s4
-; GFX11-NEXT: s_lshr_b32 s4, s2, 16
-; GFX11-NEXT: s_pack_hl_b32_b16 s1, s1, s5
-; GFX11-NEXT: s_lshr_b32 s5, s3, 16
-; GFX11-NEXT: s_lshl_b32 s2, s2, 0x80008
-; GFX11-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-NEXT: s_lshl_b32 s3, s3, 0x80008
-; GFX11-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-NEXT: s_pack_ll_b32_b16 s2, s2, s4
-; GFX11-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-NEXT: s_pack_ll_b32_b16 s3, s3, s5
-; GFX11-NEXT: s_lshr_b32 s5, s1, 16
-; GFX11-NEXT: s_lshl_b32 s0, s0, 0x80008
-; GFX11-NEXT: s_lshl_b32 s4, s4, 8
-; GFX11-NEXT: s_lshl_b32 s1, s1, 0x80008
-; GFX11-NEXT: s_lshl_b32 s5, s5, 8
-; GFX11-NEXT: v_pk_sub_u16 v0, s2, s3 clamp
-; GFX11-NEXT: s_pack_ll_b32_b16 s0, s0, s4
-; GFX11-NEXT: s_pack_ll_b32_b16 s1, s1, s5
-; GFX11-NEXT: v_pk_sub_u16 v1, s0, s1 clamp
-; GFX11-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-NEXT: v_readfirstlane_b32 s1, v1
-; GFX11-NEXT: s_and_b32 s2, s0, 0xffff
-; GFX11-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-NEXT: s_lshr_b32 s2, s2, 0x80008
-; GFX11-NEXT: s_lshr_b32 s0, s0, 8
-; GFX11-NEXT: s_and_b32 s3, s1, 0xffff
-; GFX11-NEXT: s_lshr_b32 s1, s1, 16
-; GFX11-NEXT: s_lshr_b32 s3, s3, 0x80008
-; GFX11-NEXT: s_lshr_b32 s1, s1, 8
-; GFX11-NEXT: s_pack_ll_b32_b16 s0, s2, s0
-; GFX11-NEXT: s_pack_ll_b32_b16 s1, s3, s1
-; GFX11-NEXT: s_bfe_u32 s2, s0, 0x80010
-; GFX11-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-NEXT: s_and_b32 s3, s1, 0xff
-; GFX11-NEXT: s_or_b32 s0, s0, s2
-; GFX11-NEXT: s_lshl_b32 s2, s3, 16
-; GFX11-NEXT: s_bfe_u32 s1, s1, 0x80010
-; GFX11-NEXT: s_or_b32 s0, s0, s2
-; GFX11-NEXT: s_lshl_b32 s1, s1, 24
-; GFX11-NEXT: s_or_b32 s0, s0, s1
-; GFX11-NEXT: ; return to shader part epilog
+; GFX11-TRUE16-LABEL: s_usubsat_v4i8:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 24
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
+; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
+; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s2, 16
+; GFX11-TRUE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
+; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s3, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
+; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
+; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-TRUE16-NEXT: v_pk_sub_u16 v0, s2, s3 clamp
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
+; GFX11-TRUE16-NEXT: v_pk_sub_u16 v1, s0, s1 clamp
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s1, v1
+; GFX11-TRUE16-NEXT: s_and_b32 s2, s0, 0xffff
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
+; GFX11-TRUE16-NEXT: s_and_b32 s2, s1, 0xffff
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s2, s1
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s0, 8
+; GFX11-TRUE16-NEXT: s_and_b32 s1, s1, 0xff
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s1, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 24
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX11-FAKE16-LABEL: s_usubsat_v4i8:
+; GFX11-FAKE16: ; %bb.0:
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s0, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s1, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 24
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s0, s2
+; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s0, s0, s3
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s1, s4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s2, 16
+; GFX11-FAKE16-NEXT: s_pack_hl_b32_b16 s1, s1, s5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s3, s3, s5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 0x80008
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-FAKE16-NEXT: v_pk_sub_u16 v0, s2, s3 clamp
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s4
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s5
+; GFX11-FAKE16-NEXT: v_pk_sub_u16 v1, s0, s1 clamp
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s1, v1
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s2, 0x80008
+; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xffff
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s3, 0x80008
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s2, s0
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s3, s1
+; GFX11-FAKE16-NEXT: s_bfe_u32 s2, s0, 0x80010
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s1, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s3, 16
+; GFX11-FAKE16-NEXT: s_bfe_u32 s1, s1, 0x80010
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 24
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-FAKE16-NEXT: ; return to shader part epilog
%lhs = bitcast i32 %lhs.arg to <4 x i8>
%rhs = bitcast i32 %rhs.arg to <4 x i8>
%result = call <4 x i8> @llvm.usub.sat.v4i8(<4 x i8> %lhs, <4 x i8> %rhs)
diff --git a/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll b/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll
index 9c5850a03c9ab..9db34d0f1ed1b 100644
--- a/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll
+++ b/llvm/test/CodeGen/AMDGPU/build-vector-packed-partial-undef.ll
@@ -410,25 +410,14 @@ define void @undef_lo2_v4i16(<2 x i16> %arg0) {
; GFX11-FAKE16-NEXT: ;;#ASMEND
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-TRUE16-SDAG-LABEL: undef_lo2_v4i16:
-; GFX11-TRUE16-SDAG: ; %bb.0:
-; GFX11-TRUE16-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-SDAG-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-SDAG-NEXT: ;;#ASMSTART
-; GFX11-TRUE16-SDAG-NEXT: ; use v[0:1]
-; GFX11-TRUE16-SDAG-NEXT: ;;#ASMEND
-; GFX11-TRUE16-SDAG-NEXT: s_setpc_b64 s[30:31]
-;
-; GFX11-TRUE16-GISEL-LABEL: undef_lo2_v4i16:
-; GFX11-TRUE16-GISEL: ; %bb.0:
-; GFX11-TRUE16-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-GISEL-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-GISEL-NEXT: v_mov_b16_e32 v0.h, v0.l
-; GFX11-TRUE16-GISEL-NEXT: ;;#ASMSTART
-; GFX11-TRUE16-GISEL-NEXT: ; use v[0:1]
-; GFX11-TRUE16-GISEL-NEXT: ;;#ASMEND
-; GFX11-TRUE16-GISEL-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-LABEL: undef_lo2_v4i16:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX11-TRUE16-NEXT: ;;#ASMSTART
+; GFX11-TRUE16-NEXT: ; use v[0:1]
+; GFX11-TRUE16-NEXT: ;;#ASMEND
+; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
%undef.lo = shufflevector <2 x i16> %arg0, <2 x i16> poison, <4 x i32> <i32 1, i32 1, i32 2, i32 3>
call void asm sideeffect "; use $0", "v"(<4 x i16> %undef.lo);
ret void
@@ -475,25 +464,14 @@ define void @undef_lo2_v4f16(<2 x half> %arg0) {
; GFX11-FAKE16-NEXT: ;;#ASMEND
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
-; GFX11-TRUE16-SDAG-LABEL: undef_lo2_v4f16:
-; GFX11-TRUE16-SDAG: ; %bb.0:
-; GFX11-TRUE16-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-SDAG-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-SDAG-NEXT: ;;#ASMSTART
-; GFX11-TRUE16-SDAG-NEXT: ; use v[0:1]
-; GFX11-TRUE16-SDAG-NEXT: ;;#ASMEND
-; GFX11-TRUE16-SDAG-NEXT: s_setpc_b64 s[30:31]
-;
-; GFX11-TRUE16-GISEL-LABEL: undef_lo2_v4f16:
-; GFX11-TRUE16-GISEL: ; %bb.0:
-; GFX11-TRUE16-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-GISEL-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-GISEL-NEXT: v_mov_b16_e32 v0.h, v0.l
-; GFX11-TRUE16-GISEL-NEXT: ;;#ASMSTART
-; GFX11-TRUE16-GISEL-NEXT: ; use v[0:1]
-; GFX11-TRUE16-GISEL-NEXT: ;;#ASMEND
-; GFX11-TRUE16-GISEL-NEXT: s_setpc_b64 s[30:31]
+; GFX11-TRUE16-LABEL: undef_lo2_v4f16:
+; GFX11-TRUE16: ; %bb.0:
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX11-TRUE16-NEXT: ;;#ASMSTART
+; GFX11-TRUE16-NEXT: ; use v[0:1]
+; GFX11-TRUE16-NEXT: ;;#ASMEND
+; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
%undef.lo = shufflevector <2 x half> %arg0, <2 x half> poison, <4 x i32> <i32 1, i32 1, i32 2, i32 3>
call void asm sideeffect "; use $0", "v"(<4 x half> %undef.lo);
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll b/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll
index 9105358e7bdbf..76583e806b805 100644
--- a/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll
+++ b/llvm/test/CodeGen/AMDGPU/dagcombine-fmul-sel.ll
@@ -1865,14 +1865,13 @@ define <2 x half> @fmul_select_v2f16_test3(<2 x half> %x, <2 x i32> %bool.arg1,
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v4
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: fmul_select_v2f16_test3:
@@ -2032,14 +2031,13 @@ define <2 x half> @fmul_select_v2f16_test4(<2 x half> %x, <2 x i32> %bool.arg1,
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, -1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v4
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, -1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: fmul_select_v2f16_test4:
diff --git a/llvm/test/CodeGen/AMDGPU/fma.f16.ll b/llvm/test/CodeGen/AMDGPU/fma.f16.ll
index df9e86756815f..8d1d08d1e7c68 100644
--- a/llvm/test/CodeGen/AMDGPU/fma.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fma.f16.ll
@@ -653,14 +653,12 @@ define <2 x i32> @test_D139469_v2f16(<2 x half> %arg) {
; GFX11-GISEL-TRUE16-NEXT: v_pk_mul_f16 v2, 0x291e291e, v0
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_pk_fma_f16 v0, 0x291e291e, v0, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e32 vcc_lo, 0, v2.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v0
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s0, 0, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v1.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v3.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_or_b32 s0, vcc_lo, s0
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
; GFX11-GISEL-TRUE16-NEXT: s_or_b32 s0, s1, s2
@@ -746,14 +744,12 @@ define <2 x i32> @test_D139469_v2f16(<2 x half> %arg) {
; GFX12-GISEL-TRUE16-NEXT: v_pk_mul_f16 v2, 0x291e291e, v0
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-GISEL-TRUE16-NEXT: v_pk_fma_f16 v0, 0x291e291e, v0, v1
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e32 vcc_lo, 0, v2.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v2.h
; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s0, 0, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s1, 0, v1.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v3.l
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_cmp_gt_f16_e64 s2, 0, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_or_b32 s0, vcc_lo, s0
; GFX12-GISEL-TRUE16-NEXT: s_wait_alu depctr_sa_sdst(0)
; GFX12-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0
diff --git a/llvm/test/CodeGen/AMDGPU/fmed3.ll b/llvm/test/CodeGen/AMDGPU/fmed3.ll
index e5b533ec1d60d..668347eb97004 100644
--- a/llvm/test/CodeGen/AMDGPU/fmed3.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmed3.ll
@@ -9194,22 +9194,19 @@ define <2 x half> @v_test_nnan_input_fmed3_r_i_i_v2f16_maximum_minimum(<2 x half
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, 1.0 op_sel_hi:[1,0]
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_pk_max_f16 v2, v0, 2.0 op_sel_hi:[1,0]
+; GFX11-GISEL-TRUE16-NEXT: v_pk_max_f16 v1, v0, 2.0 op_sel_hi:[1,0]
; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e32 vcc_lo, 2.0, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 2.0, v1.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v1.l, 0x7e00, v2.l, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v1.h, 0x7e00, v0.l, s0
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 2.0, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v1.l, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e32 vcc_lo, 4.0, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_pk_min_f16 v0, v1, 4.0 op_sel_hi:[1,0]
-; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 4.0, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.h, 0x7e00, v1.h, s0
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e32 vcc_lo, 4.0, v0.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v0.l, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.h, 0x7e00, v1.l, s0
+; GFX11-GISEL-TRUE16-NEXT: v_pk_min_f16 v1, v0, 4.0 op_sel_hi:[1,0]
+; GFX11-GISEL-TRUE16-NEXT: v_cmp_o_f16_e64 s0, 4.0, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.l, 0x7e00, v1.l, vcc_lo
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b16 v0.h, 0x7e00, v1.h, s0
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
%a.add = fadd nnan <2 x half> %a, splat (half 1.0)
%max = call <2 x half> @llvm.maximum.v2f16(<2 x half> %a.add, <2 x half> splat (half 2.0))
diff --git a/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll b/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll
index 8613d238d4201..95d2f07402dd4 100644
--- a/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmul-to-ldexp.ll
@@ -6793,12 +6793,11 @@ define <2 x half> @v_mul_v2f16_select_64_1(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 6, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 6, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_64_1:
@@ -6898,12 +6897,11 @@ define <2 x half> @v_mul_v2f16_select_1_64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_1_64:
@@ -7008,10 +7006,9 @@ define <2 x half> @v_mul_v2f16_select_n1_n64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_n1_n64:
@@ -7123,7 +7120,6 @@ define <2 x half> @v_mul_v2f16_select_128_64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_dual_mov_b32 v3, 0x7fff :: v_dual_add_nc_u32 v0, 6, v0
@@ -7132,7 +7128,7 @@ define <2 x half> @v_mul_v2f16_select_128_64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v1, 6, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_128_64:
@@ -7253,15 +7249,13 @@ define <2 x half> @v_mul_v2f16_select_n128_n64(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v0, 6, v0
+; GFX11-GISEL-TRUE16-NEXT: v_dual_mov_b32 v3, 0x7fff :: v_dual_add_nc_u32 v0, 6, v0
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v0, 0xffff8000, v0, v3
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v1, 6, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_n128_n64:
@@ -7372,10 +7366,9 @@ define <2 x half> @v_mul_v2f16_select_n128_n16(<2 x i32> %arg, <2 x half> %x) {
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 4, 7, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 4, 7, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v3.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_mul_v2f16_select_n128_n16:
@@ -7478,12 +7471,11 @@ define <2 x half> @v_contract_mul_add_v2f16_select_64_1(<2 x i32> %arg, <2 x hal
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 6, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 6, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7588,12 +7580,11 @@ define <2 x half> @v_contract_mul_add_v2f16_select_1_64(<2 x i32> %arg, <2 x hal
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7703,10 +7694,9 @@ define <2 x half> @v_contract_mul_add_v2f16_select_n64_n1(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 6, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 6, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7817,10 +7807,9 @@ define <2 x half> @v_contract_mul_add_v2f16_select_n1_n64(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 6, 0, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 6, 0, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -7937,7 +7926,6 @@ define <2 x half> @v_contract_mul_add_v2f16_select_128_64(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v0, 6, v0
@@ -7946,7 +7934,7 @@ define <2 x half> @v_contract_mul_add_v2f16_select_128_64(<2 x i32> %arg, <2 x h
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v0, 0xffff8000, v0, v4
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v4
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v5.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -8055,12 +8043,11 @@ define <2 x half> @v_contract_mul_add_v2f16_select_128_4(<2 x i32> %arg, <2 x ha
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 2, 7, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 2, 7, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -8176,7 +8163,6 @@ define <2 x half> @v_contract_mul_add_v2f16_select_2_4(<2 x i32> %arg, <2 x half
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_add_nc_u32_e32 v0, 2, v0
@@ -8185,7 +8171,7 @@ define <2 x half> @v_contract_mul_add_v2f16_select_2_4(<2 x i32> %arg, <2 x half
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v0, 0xffff8000, v0, v4
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v4
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v5.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -8294,12 +8280,11 @@ define <2 x half> @v_contract_mul_add_v2f16_select_4_128(<2 x i32> %arg, <2 x ha
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 7, 2, vcc_lo
; GFX11-GISEL-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v0.l
; GFX11-GISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 7, 2, vcc_lo
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_pk_add_f16 v0, v0, v3
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
index a89de01b2354c..614566a230f68 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
@@ -114,25 +114,45 @@ define amdgpu_ps <4 x float> @image_bvh_intersect_ray_a16(i32 inreg %node_ptr, f
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
; GFX12-SDAG-NEXT: ; return to shader part epilog
;
-; GFX12-GISEL-LABEL: image_bvh_intersect_ray_a16:
-; GFX12-GISEL: ; %bb.0: ; %main_body
-; GFX12-GISEL-NEXT: s_mov_b32 s20, s2
-; GFX12-GISEL-NEXT: s_mov_b32 s22, s4
-; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s4, s7, s5
-; GFX12-GISEL-NEXT: s_mov_b32 s21, s3
-; GFX12-GISEL-NEXT: s_pack_hh_b32_b16 s5, s7, s5
-; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s6, s8, s6
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v1, s21
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v3, s4
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v6, s0 :: v_dual_mov_b32 v7, s1
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s6
-; GFX12-GISEL-NEXT: s_mov_b32 s16, s9
-; GFX12-GISEL-NEXT: s_mov_b32 s17, s10
-; GFX12-GISEL-NEXT: s_mov_b32 s18, s11
-; GFX12-GISEL-NEXT: s_mov_b32 s19, s12
-; GFX12-GISEL-NEXT: image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[16:19] a16
-; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
-; GFX12-GISEL-NEXT: ; return to shader part epilog
+; GFX12-GISEL-TRUE16-LABEL: image_bvh_intersect_ray_a16:
+; GFX12-GISEL-TRUE16: ; %bb.0: ; %main_body
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s20, s2
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s22, s4
+; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s4, s7, s5
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s21, s3
+; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s8, s6
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s5, s4
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v1, s21
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v3, s4
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v6, s0 :: v_dual_mov_b32 v7, s1
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s6
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s16, s9
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s17, s10
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s18, s11
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s19, s12
+; GFX12-GISEL-TRUE16-NEXT: image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[16:19] a16
+; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-FAKE16-LABEL: image_bvh_intersect_ray_a16:
+; GFX12-GISEL-FAKE16: ; %bb.0: ; %main_body
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s20, s2
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s22, s4
+; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s7, s5
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s21, s3
+; GFX12-GISEL-FAKE16-NEXT: s_pack_hh_b32_b16 s5, s7, s5
+; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s6, s8, s6
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v1, s21
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v3, s4
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v6, s0 :: v_dual_mov_b32 v7, s1
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s6
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s16, s9
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s17, s10
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s18, s11
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s19, s12
+; GFX12-GISEL-FAKE16-NEXT: image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[16:19] a16
+; GFX12-GISEL-FAKE16-NEXT: s_wait_bvhcnt 0x0
+; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
main_body:
%v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i32.v4f16(i32 %node_ptr, float %ray_extent, <3 x float> %ray_origin, <3 x half> %ray_dir, <3 x half> %ray_inv_dir, <4 x i32> %tdescr)
%r = bitcast <4 x i32> %v to <4 x float>
@@ -236,26 +256,47 @@ define amdgpu_ps <4 x float> @image_bvh64_intersect_ray_a16(i64 inreg %node_ptr,
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
; GFX12-SDAG-NEXT: ; return to shader part epilog
;
-; GFX12-GISEL-LABEL: image_bvh64_intersect_ray_a16:
-; GFX12-GISEL: ; %bb.0: ; %main_body
-; GFX12-GISEL-NEXT: s_mov_b32 s20, s3
-; GFX12-GISEL-NEXT: s_mov_b32 s21, s4
-; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s4, s8, s6
-; GFX12-GISEL-NEXT: s_mov_b32 s22, s5
-; GFX12-GISEL-NEXT: s_pack_hh_b32_b16 s5, s8, s6
-; GFX12-GISEL-NEXT: s_pack_ll_b32_b16 s6, s9, s7
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v3, s4
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v1, s21
-; GFX12-GISEL-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v5, s6
-; GFX12-GISEL-NEXT: v_mov_b32_e32 v4, s5
-; GFX12-GISEL-NEXT: s_mov_b32 s16, s10
-; GFX12-GISEL-NEXT: s_mov_b32 s17, s11
-; GFX12-GISEL-NEXT: s_mov_b32 s18, s12
-; GFX12-GISEL-NEXT: s_mov_b32 s19, s13
-; GFX12-GISEL-NEXT: image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[16:19] a16
-; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
-; GFX12-GISEL-NEXT: ; return to shader part epilog
+; GFX12-GISEL-TRUE16-LABEL: image_bvh64_intersect_ray_a16:
+; GFX12-GISEL-TRUE16: ; %bb.0: ; %main_body
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s21, s4
+; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s4, s8, s6
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s20, s3
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s22, s5
+; GFX12-GISEL-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s9, s7
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s5, s4
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v3, s4
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v1, s21
+; GFX12-GISEL-TRUE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v5, s6
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b32_e32 v4, s5
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s16, s10
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s17, s11
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s18, s12
+; GFX12-GISEL-TRUE16-NEXT: s_mov_b32 s19, s13
+; GFX12-GISEL-TRUE16-NEXT: image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[16:19] a16
+; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
+; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
+; GFX12-GISEL-FAKE16-LABEL: image_bvh64_intersect_ray_a16:
+; GFX12-GISEL-FAKE16: ; %bb.0: ; %main_body
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s20, s3
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s21, s4
+; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s8, s6
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s22, s5
+; GFX12-GISEL-FAKE16-NEXT: s_pack_hh_b32_b16 s5, s8, s6
+; GFX12-GISEL-FAKE16-NEXT: s_pack_ll_b32_b16 s6, s9, s7
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v0, s20 :: v_dual_mov_b32 v3, s4
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v7, s1 :: v_dual_mov_b32 v6, s0
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v1, s21
+; GFX12-GISEL-FAKE16-NEXT: v_dual_mov_b32 v2, s22 :: v_dual_mov_b32 v5, s6
+; GFX12-GISEL-FAKE16-NEXT: v_mov_b32_e32 v4, s5
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s16, s10
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s17, s11
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s18, s12
+; GFX12-GISEL-FAKE16-NEXT: s_mov_b32 s19, s13
+; GFX12-GISEL-FAKE16-NEXT: image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[16:19] a16
+; GFX12-GISEL-FAKE16-NEXT: s_wait_bvhcnt 0x0
+; GFX12-GISEL-FAKE16-NEXT: ; return to shader part epilog
main_body:
%v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i64.v4f16(i64 %node_ptr, float %ray_extent, <3 x float> %ray_origin, <3 x half> %ray_dir, <3 x half> %ray_inv_dir, <4 x i32> %tdescr)
%r = bitcast <4 x i32> %v to <4 x float>
@@ -891,7 +932,5 @@ declare i32 @llvm.amdgcn.workitem.id.x()
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX11-FAKE16: {{.*}}
; GFX11-TRUE16: {{.*}}
-; GFX12-GISEL-FAKE16: {{.*}}
-; GFX12-GISEL-TRUE16: {{.*}}
; GFX12-SDAG-FAKE16: {{.*}}
; GFX12-SDAG-TRUE16: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll b/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll
index 6510824ec3ea6..7300e4227925f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.frexp.ll
@@ -535,15 +535,13 @@ define { <2 x half>, <2 x i32> } @test_frexp_v2f16_v2i32(<2 x half> %a) {
; GFX11-GISEL-TRUE16-LABEL: test_frexp_v2f16_v2i32:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.l
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.l
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v3.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v2, 0, 16
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v3, 0, 16
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
+; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v2, 0, 16
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32:
@@ -568,15 +566,13 @@ define { <2 x half>, <2 x i32> } @test_frexp_v2f16_v2i32(<2 x half> %a) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.l
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.l
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v2.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v3.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v2, 0, 16
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v3, 0, 16
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
+; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v2, v2, 0, 16
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32:
@@ -709,10 +705,8 @@ define <2 x half> @test_frexp_v2f16_v2i32_only_use_fract(<2 x half> %a) {
; GFX11-GISEL-TRUE16-LABEL: test_frexp_v2f16_v2i32_only_use_fract:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32_only_use_fract:
@@ -732,10 +726,8 @@ define <2 x half> @test_frexp_v2f16_v2i32_only_use_fract(<2 x half> %a) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_mant_f16_e32 v0.h, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-FAKE16-LABEL: test_frexp_v2f16_v2i32_only_use_fract:
@@ -880,12 +872,10 @@ define <2 x i32> @test_frexp_v2f16_v2i32_only_use_exp(<2 x half> %a) {
; GFX11-GISEL-TRUE16-LABEL: test_frexp_v2f16_v2i32_only_use_exp:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v0.l, v0.l
+; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v0, v0, 0, 16
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -908,12 +898,10 @@ define <2 x i32> @test_frexp_v2f16_v2i32_only_use_exp(<2 x half> %a) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v0.l, v0.l
+; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-GISEL-TRUE16-NEXT: v_frexp_exp_i16_f16_e32 v1.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v0, v0, 0, 16
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX12-GISEL-TRUE16-NEXT: v_bfe_i32 v1, v1, 0, 16
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
index 3b5763b784184..cc5f6f842625c 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.is.fpclass.f16.ll
@@ -1446,10 +1446,11 @@ define <2 x i1> @isnan_v2f16(<2 x half> %x) nounwind {
; GFX11GLISEL-TRUE16: ; %bb.0:
; GFX11GLISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v1, 3 :: v_dual_mov_b32 v2, 3
-; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v0
; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.l, v1.l
-; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v3.l, v2.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v3.l, v2.l
+; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.h, v3.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v0, v2
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX11GLISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -1608,13 +1609,14 @@ define <3 x i1> @isnan_v3f16(<3 x half> %x) nounwind {
; GFX11GLISEL-TRUE16: ; %bb.0:
; GFX11GLISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v2, 3 :: v_dual_mov_b32 v3, 3
-; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v0
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v4, 3
; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.l, v2.l
-; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v5.l, v3.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.l
+; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v4, 0, 1, vcc_lo
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.h, v3.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v0, v4
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v4, 3
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v4.l
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v5.l
; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v1, v3
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
; GFX11GLISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -1803,19 +1805,18 @@ define <4 x i1> @isnan_v4f16(<4 x half> %x) nounwind {
; GFX11GLISEL-TRUE16: ; %bb.0:
; GFX11GLISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v2, 3 :: v_dual_mov_b32 v3, 3
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v4, 3
-; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v0
-; GFX11GLISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v1
+; GFX11GLISEL-TRUE16-NEXT: v_dual_mov_b32 v4, 3 :: v_dual_mov_b32 v5, 3
; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.l, v2.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.l
-; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v7.l, v3.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v6.l, v4.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b16_e32 v7.l, v5.l
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v4, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v6, 3
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v5.l
-; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v0.h, v3.l
+; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v5, 0, 1, vcc_lo
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.l, v6.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v0, v4
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo
-; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v8.l, v6.l
+; GFX11GLISEL-TRUE16-NEXT: v_cmp_class_f16_e32 vcc_lo, v1.h, v7.l
+; GFX11GLISEL-TRUE16-NEXT: v_mov_b32_e32 v1, v5
; GFX11GLISEL-TRUE16-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo
; GFX11GLISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll b/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll
index dcda6583527e4..43cc632d3708b 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.ldexp.ll
@@ -512,13 +512,12 @@ define <2 x half> @test_ldexp_v2f16_v2i32(<2 x half> %a, <2 x i32> %b) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v3, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v1, v3
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v3
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v2f16_v2i32:
@@ -626,11 +625,8 @@ define <2 x half> @test_ldexp_v2f16_v2i16(<2 x half> %a, <2 x i16> %b) {
; GFX11-GISEL-TRUE16-LABEL: test_ldexp_v2f16_v2i16:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v2f16_v2i16:
@@ -775,14 +771,13 @@ define <3 x half> @test_ldexp_v3f16_v3i32(<3 x half> %a, <3 x i32> %b) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v5, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v5
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v3, v5
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v5
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v6.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v4.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -911,12 +906,9 @@ define <3 x half> @test_ldexp_v3f16_v3i16(<3 x half> %a, <3 x i16> %b) {
; GFX11-GISEL-TRUE16-LABEL: test_ldexp_v3f16_v3i16:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v3f16_v3i16:
@@ -1096,19 +1088,17 @@ define <4 x half> @test_ldexp_v4f16_v4i32(<4 x half> %a, <4 x i32> %b) {
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v6, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v1
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v2, 0xffff8000, v2, v6
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v3, v6
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v6
; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v5, v6
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v7.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v4.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v8.l, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v1.h, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v4f16_v4i32:
@@ -1267,15 +1257,10 @@ define <4 x half> @test_ldexp_v4f16_v4i16(<4 x half> %a, <4 x i16> %b) {
; GFX11-GISEL-TRUE16-LABEL: test_ldexp_v4f16_v4i16:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v0.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v1.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v4.l, v6.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v5.l, v7.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v1.h, v3.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v4f16_v4i16:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log.ll b/llvm/test/CodeGen/AMDGPU/llvm.log.ll
index 0c5adcbb8c4df..7903ae93d770c 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log.ll
@@ -6722,10 +6722,8 @@ define <2 x half> @v_log_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v2f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -6862,10 +6860,9 @@ define <2 x half> @v_log_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7006,10 +7003,9 @@ define <2 x half> @v_log_fneg_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_or_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7151,10 +7147,9 @@ define <2 x half> @v_log_fneg_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7280,10 +7275,8 @@ define <2 x half> @v_log_v2f16_fast(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v2f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
@@ -7393,15 +7386,14 @@ define <3 x half> @v_log_v3f16(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v3f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v3f16:
@@ -7511,15 +7503,14 @@ define <3 x half> @v_log_v3f16_fast(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v3f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v3f16_fast:
@@ -7707,19 +7698,16 @@ define <4 x half> @v_log_v4f16(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v4f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v1.h
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v1.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v4f16:
@@ -7912,19 +7900,16 @@ define <4 x half> @v_log_v4f16_fast(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log_v4f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x398c, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x398c, v1.h
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x398c, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x398c, v1.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log_v4f16_fast:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log10.ll b/llvm/test/CodeGen/AMDGPU/llvm.log10.ll
index e2fcb18ee3217..478580ff8ec0a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log10.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log10.ll
@@ -6722,10 +6722,8 @@ define <2 x half> @v_log10_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v2f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -6862,10 +6860,9 @@ define <2 x half> @v_log10_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7006,10 +7003,9 @@ define <2 x half> @v_log10_fneg_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_or_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7151,10 +7147,9 @@ define <2 x half> @v_log10_fneg_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7280,10 +7275,8 @@ define <2 x half> @v_log10_v2f16_fast(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v2f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
@@ -7393,15 +7386,14 @@ define <3 x half> @v_log10_v3f16(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v3f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v3f16:
@@ -7511,15 +7503,14 @@ define <3 x half> @v_log10_v3f16_fast(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v3f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v3f16_fast:
@@ -7707,19 +7698,16 @@ define <4 x half> @v_log10_v4f16(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v4f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v1.h
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v1.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v4f16:
@@ -7912,19 +7900,16 @@ define <4 x half> @v_log10_v4f16_fast(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log10_v4f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v2.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v2.l, v3.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, 0x34d1, v0.l
; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(TRANS32_DEP_3)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v0.h
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt_depctr depctr_va_vdst(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.h, 0x34d1, v1.h
-; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v2.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.l, 0x34d1, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_mul_f16_e32 v1.h, 0x34d1, v1.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log10_v4f16_fast:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log2.ll b/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
index 0abf90907d2f9..8401e05b39c19 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
@@ -4094,10 +4094,8 @@ define <2 x half> @v_log2_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v2f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v2f16:
@@ -4210,10 +4208,9 @@ define <2 x half> @v_log2_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_fabs_v2f16:
@@ -4314,10 +4311,9 @@ define <2 x half> @v_log2_fneg_fabs_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_or_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_fneg_fabs_v2f16:
@@ -4419,10 +4415,9 @@ define <2 x half> @v_log2_fneg_v2f16(<2 x half> %in) {
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1100-GISEL-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
+; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_fneg_v2f16:
@@ -4519,10 +4514,8 @@ define <2 x half> @v_log2_v2f16_fast(<2 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v2f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v1.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v2f16_fast:
@@ -4626,11 +4619,9 @@ define <3 x half> @v_log2_v3f16(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v3f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v3f16:
@@ -4735,11 +4726,9 @@ define <3 x half> @v_log2_v3f16_fast(<3 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v3f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v3f16_fast:
@@ -4885,13 +4874,10 @@ define <4 x half> @v_log2_v4f16(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v4f16:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v3.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v4f16:
@@ -5040,13 +5026,10 @@ define <4 x half> @v_log2_v4f16_fast(<4 x half> %in) {
; GFX1100-GISEL-TRUE16-LABEL: v_log2_v4f16_fast:
; GFX1100-GISEL-TRUE16: ; %bb.0:
; GFX1100-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX1100-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.l, v0.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v0.h
; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.l, v1.l
-; GFX1100-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v0.h, v2.l
-; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v3.l
+; GFX1100-GISEL-TRUE16-NEXT: v_log_f16_e32 v1.h, v1.h
; GFX1100-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX1100-GISEL-FAKE16-LABEL: v_log2_v4f16_fast:
diff --git a/llvm/test/CodeGen/AMDGPU/roundeven.ll b/llvm/test/CodeGen/AMDGPU/roundeven.ll
index fd083d6eca49c..8920bfbd3b9dc 100644
--- a/llvm/test/CodeGen/AMDGPU/roundeven.ll
+++ b/llvm/test/CodeGen/AMDGPU/roundeven.ll
@@ -482,9 +482,8 @@ define <2 x half> @v_roundeven_v2f16(<2 x half> %x) {
; GFX11-TRUE16-LABEL: v_roundeven_v2f16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v1.l
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v0.h
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_roundeven_v2f16:
@@ -629,9 +628,8 @@ define <2 x half> @v_roundeven_v2f16_fneg(<2 x half> %x) {
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v1.l
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v0.h
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_roundeven_v2f16_fneg:
@@ -801,12 +799,10 @@ define <4 x half> @v_roundeven_v4f16(<4 x half> %x) {
; GFX11-TRUE16-LABEL: v_roundeven_v4f16:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.l, v0.l
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v0.h
; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v1.l, v1.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v0.h, v2.l
-; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v1.h, v3.l
+; GFX11-TRUE16-NEXT: v_rndne_f16_e32 v1.h, v1.h
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: v_roundeven_v4f16:
diff --git a/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll b/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll
index b2d9965bf1fed..0787e0d0551fc 100644
--- a/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/strict_fsub.f16.ll
@@ -586,11 +586,9 @@ define <3 x half> @v_constained_fsub_v3f16_fpexcept_strict(<3 x half> %x, <3 x h
; GFX11-GISEL-TRUE16-LABEL: v_constained_fsub_v3f16_fpexcept_strict:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v2
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.l, v0.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.l, v1.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v4.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_constained_fsub_v3f16_fpexcept_strict:
@@ -766,14 +764,10 @@ define <4 x half> @v_constained_fsub_v4f16_fpexcept_strict(<4 x half> %x, <4 x h
; GFX11-GISEL-TRUE16-LABEL: v_constained_fsub_v4f16_fpexcept_strict:
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.l, v0.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v0.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.l, v1.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v0.h, v4.l, v6.l
-; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.h, v5.l, v7.l
+; GFX11-GISEL-TRUE16-NEXT: v_sub_f16_e32 v1.h, v1.h, v3.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: v_constained_fsub_v4f16_fpexcept_strict:
diff --git a/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll b/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll
index fba9d6c0e6488..fd313a76fc675 100644
--- a/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/strict_ldexp.f16.ll
@@ -178,13 +178,12 @@ define <2 x half> @test_ldexp_v2f16_v2i32(ptr addrspace(1) %out, <2 x half> %a,
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v0, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v3, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v0
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v3, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v3, 0xffff8000, v4, v0
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v1.l, v4.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v3.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v2f16_v2i32:
@@ -300,15 +299,14 @@ define <3 x half> @test_ldexp_v3f16_v3i32(ptr addrspace(1) %out, <3 x half> %a,
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v0, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v5, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v6, 0xffff8000, v6, v0
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v4.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v4, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v5, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v6, v0
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v1.l, v5.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v6.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v4.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v3f16_v3i32:
@@ -447,19 +445,17 @@ define <4 x half> @test_ldexp_v4f16_v4i32(ptr addrspace(1) %out, <4 x half> %a,
; GFX11-GISEL-TRUE16: ; %bb.0:
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-TRUE16-NEXT: v_mov_b32_e32 v0, 0x7fff
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v3
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v4, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v5, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v6, 0xffff8000, v6, v0
-; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v7, 0xffff8000, v7, v0
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v4.l
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v1, 0xffff8000, v4, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v4, 0xffff8000, v5, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v5, 0xffff8000, v6, v0
+; GFX11-GISEL-TRUE16-NEXT: v_med3_i32 v6, 0xffff8000, v7, v0
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.l, v2.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v1.l, v5.l
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v6.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v0.h, v2.h, v4.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.l, v3.l, v5.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v8.l, v7.l
+; GFX11-GISEL-TRUE16-NEXT: v_ldexp_f16_e32 v1.h, v3.h, v6.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-GISEL-FAKE16-LABEL: test_ldexp_v4f16_v4i32:
diff --git a/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll b/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll
index 680ae989e6e21..99b6ab7a6401b 100644
--- a/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/v_sat_pk_u8_i16.ll
@@ -1615,10 +1615,9 @@ define i16 @basic_smax_smin_vec_input(<2 x i16> %src) {
; GISEL-GFX11-TRUE16-NEXT: v_pk_min_i16 v0, 0xff00ff, v0
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX11-TRUE16-NEXT: v_pk_max_i16 v1, 0, v0
-; GISEL-GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
+; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
+; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
-; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
; GISEL-GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GISEL-GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
@@ -1649,10 +1648,9 @@ define i16 @basic_smax_smin_vec_input(<2 x i16> %src) {
; GISEL-GFX12-TRUE16-NEXT: v_pk_min_i16 v0, 0xff00ff, v0
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX12-TRUE16-NEXT: v_pk_max_i16 v1, 0, v0
-; GISEL-GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
+; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.h
+; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
-; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
; GISEL-GFX12-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GISEL-GFX12-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
@@ -1790,13 +1788,9 @@ define i16 @basic_smax_smin_vec_input_rev(<2 x i16> %src) {
; GISEL-GFX11-TRUE16-NEXT: v_pk_max_i16 v0, 0, v0
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX11-TRUE16-NEXT: v_pk_min_i16 v1, 0xff00ff, v0
-; GISEL-GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
-; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
-; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GISEL-GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
-; GISEL-GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GISEL-GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
; GISEL-GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GISEL-GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GISEL-GFX11-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
; GISEL-GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX11-FAKE16-LABEL: basic_smax_smin_vec_input_rev:
@@ -1824,13 +1818,9 @@ define i16 @basic_smax_smin_vec_input_rev(<2 x i16> %src) {
; GISEL-GFX12-TRUE16-NEXT: v_pk_max_i16 v0, 0, v0
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GISEL-GFX12-TRUE16-NEXT: v_pk_min_i16 v1, 0xff00ff, v0
-; GISEL-GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, 16, v1
-; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
-; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GISEL-GFX12-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v0.l
-; GISEL-GFX12-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GISEL-GFX12-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v1.h
; GISEL-GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GISEL-GFX12-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
+; GISEL-GFX12-TRUE16-NEXT: v_or_b16 v0.l, v1.l, v0.l
; GISEL-GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GISEL-GFX12-FAKE16-LABEL: basic_smax_smin_vec_input_rev:
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
index c86c1a32e4cf7..0ec4c18a070fe 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fadd.ll
@@ -239,10 +239,9 @@ define half @test_vector_reduce_fadd_v3half(half %sp, <3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fadd_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -290,10 +289,9 @@ define half @test_vector_reduce_fadd_v3half(half %sp, <3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v3.l
+; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v2.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll
index f9fc6e9b1eea3..44b8b8bcb9ae8 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmax.ll
@@ -107,10 +107,9 @@ define half @test_vector_reduce_fmax_v2half(<2 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v2half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -158,10 +157,9 @@ define half @test_vector_reduce_fmax_v2half(<2 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -288,13 +286,11 @@ define half @test_vector_reduce_fmax_v3half(<3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v2.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -350,13 +346,11 @@ define half @test_vector_reduce_fmax_v3half(<3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v2.l, v2.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -513,16 +507,13 @@ define half @test_vector_reduce_fmax_v4half(<4 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v4half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -581,16 +572,13 @@ define half @test_vector_reduce_fmax_v4half(<4 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -838,20 +826,16 @@ define half @test_vector_reduce_fmax_v8half(<8 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v8half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v4.l, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v5.l, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v6.l, v6.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v7.l, v7.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v3.h, v3.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.h
@@ -934,20 +918,16 @@ define half @test_vector_reduce_fmax_v8half(<8 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v4.l, v4.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v5.l, v5.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v6.l, v6.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v7.l, v7.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v3.h, v3.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.h
@@ -1376,33 +1356,26 @@ define half @test_vector_reduce_fmax_v16half(<16 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmax_v16half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v8.l, v8.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v9.l, v9.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v10.l, v10.l
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.h
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v11.l, v11.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v3.h, v3.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v12.l, v12.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v4.h, v4.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v5.l, v5.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v13.l, v13.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v5.h, v5.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.h, v6.l, v6.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v8.l, v8.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v6.h, v6.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.h, v7.l, v7.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v9.l, v9.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v7.h, v7.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v2.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.h, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v3.h, v4.l
@@ -1527,33 +1500,26 @@ define half @test_vector_reduce_fmax_v16half(<16 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v8.l, v8.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v9.l, v9.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v10.l, v10.l
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.h
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v11.l, v11.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v3.h, v3.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v12.l, v12.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v4.h, v4.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v5.l, v5.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v13.l, v13.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v5.h, v5.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v6.l, v6.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v8.l, v8.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v6.h, v6.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.h, v7.l, v7.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v9.l, v9.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v7.h, v7.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v2.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.h, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v3.h, v4.l
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll
index 4db8e38f94b2c..ed5c910def3d6 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmin.ll
@@ -107,10 +107,9 @@ define half @test_vector_reduce_fmin_v2half(<2 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v2half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -158,10 +157,9 @@ define half @test_vector_reduce_fmin_v2half(<2 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -288,13 +286,11 @@ define half @test_vector_reduce_fmin_v3half(<3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v2.l, v2.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -350,13 +346,11 @@ define half @test_vector_reduce_fmin_v3half(<3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v2.l, v2.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -513,16 +507,13 @@ define half @test_vector_reduce_fmin_v4half(<4 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v4half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v0.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v1.l, v1.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -581,16 +572,13 @@ define half @test_vector_reduce_fmin_v4half(<4 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v0.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v1.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -838,20 +826,16 @@ define half @test_vector_reduce_fmin_v8half(<8 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v8half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v4.l, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v5.l, v5.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v6.l, v6.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v7.l, v7.l
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v0.h, v1.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v3.h, v3.h
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v1.l, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.l, v2.l, v2.h
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.h, v3.l, v3.h
@@ -934,20 +918,16 @@ define half @test_vector_reduce_fmin_v8half(<8 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v4.l, v4.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v5.l, v5.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v6.l, v6.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v7.l, v7.l
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v0.h, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v3.h, v3.h
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v1.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v2.l, v2.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.h, v3.l, v3.h
@@ -1376,33 +1356,26 @@ define half @test_vector_reduce_fmin_v16half(<16 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmin_v16half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v8.l, v8.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v0.h, v0.h, v0.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.l, v1.l, v1.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v9.l, v9.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v1.h, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v2.l, v2.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v10.l, v10.l
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v2.h, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.l, v0.l, v0.h
+; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v0.h, v1.l, v1.h
-; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.l, v2.l, v2.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v1.h, v3.l, v3.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v11.l, v11.l
+; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.l, v2.l, v2.h
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.l, v3.h, v3.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v2.h, v4.l, v4.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v12.l, v12.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.l, v4.h, v4.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v3.h, v5.l, v5.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v13.l, v13.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.l, v5.h, v5.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v4.h, v6.l, v6.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v8.l, v8.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.l, v6.h, v6.h
; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v5.h, v7.l, v7.l
-; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v9.l, v9.l
+; GFX11-GISEL-TRUE16-NEXT: v_max_f16_e32 v6.l, v7.h, v7.h
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v1.h, v1.h, v2.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v2.l, v2.h, v3.l
; GFX11-GISEL-TRUE16-NEXT: v_min_f16_e32 v2.h, v3.h, v4.l
@@ -1527,33 +1500,26 @@ define half @test_vector_reduce_fmin_v16half(<16 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.l, v0.l, v0.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v8.l, v8.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v0.h, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v1.l, v1.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v9.l, v9.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v1.h, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v2.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v10.l, v10.l
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v7
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v2.h, v2.h
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v0.h, v1.l, v1.h
-; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v2.l, v2.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v1.h, v3.l, v3.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v11.l, v11.l
+; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v2.l, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.l, v3.h, v3.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v2.h, v4.l, v4.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v12.l, v12.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.l, v4.h, v4.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v3.h, v5.l, v5.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v13.l, v13.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.l, v5.h, v5.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v6.l, v6.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v8.l, v8.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v6.h, v6.h
; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v5.h, v7.l, v7.l
-; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v9.l, v9.l
+; GFX12-GISEL-TRUE16-NEXT: v_max_num_f16_e32 v6.l, v7.h, v7.h
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v1.h, v1.h, v2.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v2.l, v2.h, v3.l
; GFX12-GISEL-TRUE16-NEXT: v_min_num_f16_e32 v2.h, v3.h, v4.l
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll
index f85035fb3ac21..63e42e1e8a320 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fminimum.ll
@@ -100,9 +100,7 @@ define half @test_vector_reduce_fminimum_v2half(<2 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v1.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX12-GISEL-FAKE16-LABEL: test_vector_reduce_fminimum_v2half:
@@ -243,9 +241,8 @@ define half @test_vector_reduce_fminimum_v3half(<3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v2.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -410,11 +407,8 @@ define half @test_vector_reduce_fminimum_v4half(<4 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v2.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v3.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
@@ -681,16 +675,10 @@ define half @test_vector_reduce_fminimum_v8half(<8 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v4, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v2
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v3
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v4.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v5.l
-; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v6.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v7.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v3.h
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
@@ -1181,22 +1169,14 @@ define half @test_vector_reduce_fminimum_v16half(<16 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v8, 16, v0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v1
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v10, 16, v2
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v11, 16, v3
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v12, 16, v4
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 16, v5
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v14, 16, v6
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v15, 16, v7
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v8.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v9.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v10.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v11.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.l, v4.l, v12.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.h, v5.l, v13.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.l, v6.l, v14.l
-; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.h, v7.l, v15.l
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v2.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.h, v3.l, v3.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.l, v4.l, v4.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v2.h, v5.l, v5.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.l, v6.l, v6.h
+; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v3.h, v7.l, v7.h
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.l, v0.l, v0.h
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v0.h, v1.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_minimum_f16 v1.l, v2.l, v2.h
diff --git a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll
index 6e0548c35552c..57dc288bf6dcd 100644
--- a/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector-reduce-fmul.ll
@@ -239,10 +239,9 @@ define half @test_vector_reduce_fmul_v3half(half %sp, <3 x half> %v) {
; GFX11-GISEL-TRUE16-LABEL: test_vector_reduce_fmul_v3half:
; GFX11-GISEL-TRUE16: ; %bb.0: ; %entry
; GFX11-GISEL-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.l
; GFX11-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v3.l
+; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.h
; GFX11-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v2.l
; GFX11-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -290,10 +289,9 @@ define half @test_vector_reduce_fmul_v3half(half %sp, <3 x half> %v) {
; GFX12-GISEL-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-GISEL-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 16, v1
; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.l
; GFX12-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v3.l
+; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v1.h
; GFX12-GISEL-TRUE16-NEXT: v_mul_f16_e32 v0.l, v0.l, v2.l
; GFX12-GISEL-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
More information about the llvm-commits
mailing list