[llvm] [AMDGPU][True16] Legalize S16 for stores & loads in GlobalISel when True16 is enabled (PR #176963)
Domenic Nutile via llvm-commits
llvm-commits at lists.llvm.org
Mon May 4 09:13:56 PDT 2026
https://github.com/saxlungs updated https://github.com/llvm/llvm-project/pull/176963
>From 39bee23e086244590da9590fb3fdd97118abdf4f Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Tue, 21 Apr 2026 12:38:59 -0400
Subject: [PATCH 1/6] [AMDGPU][True16] Legalize S16 for G_STORE/G_LOAD when
True16 is enabled
Signed-off-by: Domenic Nutile <domenic.nutile at gmail.com>
---
.../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 15 +-
.../AMDGPU/AMDGPURegBankLegalizeRules.cpp | 2 +-
.../CodeGen/AMDGPU/GlobalISel/flat-scratch.ll | 174 +-
.../AMDGPU/GlobalISel/insertelement.i8.ll | 488 +-
.../GlobalISel/legalize-load-constant.mir | 2466 +++-
.../AMDGPU/GlobalISel/legalize-load-flat.mir | 2400 ++--
.../GlobalISel/legalize-load-global.mir | 10474 +++++++++++++++-
.../AMDGPU/GlobalISel/legalize-load-local.mir | 4052 +++++-
.../GlobalISel/legalize-load-private.mir | 2020 ++-
.../GlobalISel/legalize-store-global.mir | 2430 +++-
.../CodeGen/AMDGPU/GlobalISel/load-d16.ll | 2 +
.../AMDGPU/GlobalISel/store-local.128.ll | 64 +-
.../AMDGPU/GlobalISel/store-local.96.ll | 48 +-
llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll | 118 +-
llvm/test/CodeGen/AMDGPU/global-saddr-load.ll | 129 +-
15 files changed, 22694 insertions(+), 2188 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 4e956db103188..b3afa80675a07 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -464,8 +464,8 @@ 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 allow extloads to up to 32 bits.
+ if (MemSize != RegSize && RegSize > 32)
return false;
if (MemSize > maxSizeForAddrSpace(ST, AS, IsLoad,
@@ -544,7 +544,7 @@ 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);
}
@@ -1573,6 +1573,15 @@ 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);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
index 977dd40af2741..b5a82eb25b82b 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
@@ -1101,7 +1101,7 @@ RegBankLegalizeRules::RegBankLegalizeRules(const GCNSubtarget &_ST,
.Any({{DivB64, P5}, {{VgprB64}, {VgprP5}}})
.Any({{DivB96, P5}, {{VgprB96}, {VgprP5}}})
.Any({{DivB128, P5}, {{VgprB128}, {VgprP5}}})
-
+
.Any({{DivS32, Ptr128}, {{Vgpr32}, {VgprPtr128}}});
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
index a9876ce5d99bc..fc0f84b0594dc 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
@@ -3,13 +3,13 @@
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=+enable-flat-scratch < %s | FileCheck -check-prefix=GFX10 %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds < %s | FileCheck -check-prefix=GFX942 %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds < %s | FileCheck -check-prefix=GFX11 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds < %s | FileCheck -check-prefix=GFX12 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=+real-true16 < %s | FileCheck -check-prefix=GFX12 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -global-isel -new-reg-bank-select -mattr=-unaligned-access-mode -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX9 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 -global-isel -new-reg-bank-select -mattr=-unaligned-access-mode -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX10 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -global-isel -new-reg-bank-select -mattr=-unaligned-access-mode -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds < %s | FileCheck -check-prefixes=UNALIGNED_GFX942 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -global-isel -new-reg-bank-select -mattr=-unaligned-access-mode -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds < %s | FileCheck -check-prefixes=UNALIGNED_GFX11 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -global-isel -new-reg-bank-select -mattr=-unaligned-access-mode -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds < %s | FileCheck -check-prefixes=UNALIGNED_GFX12 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX10 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode < %s | FileCheck -check-prefixes=UNALIGNED_GFX942 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode < %s | FileCheck -check-prefixes=UNALIGNED_GFX11 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=+real-true16,-unaligned-access-mode < %s | FileCheck -check-prefixes=UNALIGNED_GFX12 %s
define amdgpu_kernel void @store_load_sindex_kernel(i32 %idx) {
; GFX9-LABEL: store_load_sindex_kernel:
@@ -2689,22 +2689,23 @@ 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)
@@ -2731,23 +2732,24 @@ define void @store_load_i64_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 15 :: v_dual_mov_b32 v2, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 15
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
@@ -3028,32 +3030,33 @@ 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)
@@ -3088,33 +3091,34 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 2
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 1
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 2
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v1, 3
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 3
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off offset:8 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:9 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:10 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:11 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
@@ -3458,41 +3462,42 @@ 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)
@@ -3535,42 +3540,43 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 2
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 1
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 2
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v1, 3
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 4
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 3
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 4
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off offset:8 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:9 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:10 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:11 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:12 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:12 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:13 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:13 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:14 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:14 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:15 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:15 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/insertelement.i8.ll
index 30d6c871115e7..3864463e2c2b6 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 -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji < %s | FileCheck --check-prefix=GFX8 %s
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=hawaii < %s | FileCheck --check-prefix=GFX7 %s
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck --check-prefix=GFX10 %s
-; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX11 %s
+; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck --check-prefixes=GFX11,GFX11-TRUE16 %s
+; RUN: llc -global-isel -new-reg-bank-select -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:
@@ -85,24 +86,45 @@ 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_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
-; GFX11-NEXT: s_mov_b32 m0, s5
-; GFX11-NEXT: global_load_u16 v0, v0, s[2:3]
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_readfirstlane_b32 s0, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: s_lshr_b32 s1, s0, 8
-; GFX11-NEXT: s_movreld_b32 s0, s4
-; GFX11-NEXT: s_and_b32 s1, s1, 0xff
-; GFX11-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-NEXT: s_lshl_b32 s1, s1, 8
-; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT: s_or_b32 s0, s0, s1
-; GFX11-NEXT: v_mov_b32_e32 v2, s0
-; 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_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 m0, s5
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: global_load_u16 v0, v0, s[2:3]
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-TRUE16-NEXT: s_movreld_b32 s0, s4
+; GFX11-TRUE16-NEXT: s_and_b32 s1, s1, 0xff
+; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
+; 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_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 m0, s5
+; GFX11-FAKE16-NEXT: global_load_u16 v0, v0, s[2:3]
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
+; GFX11-FAKE16-NEXT: s_movreld_b32 s0, s4
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, s0
+; 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
@@ -178,23 +200,40 @@ define amdgpu_ps void @insertelement_v_v2i8_s_s(ptr addrspace(1) %ptr, i8 inreg
; GFX10-NEXT: global_store_short v[1:2], v0, 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: s_mov_b32 m0, s3
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
-; GFX11-NEXT: v_movreld_b32_e32 v0, s2
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; 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 v1, v[0:1], off
+; GFX11-TRUE16-NEXT: s_mov_b32 m0, s3
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v2.l, 8, v1.l
+; GFX11-TRUE16-NEXT: v_movreld_b32_e32 v1, s2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; 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: s_mov_b32 m0, s3
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-FAKE16-NEXT: v_movreld_b32_e32 v0, s2
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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
@@ -283,27 +322,49 @@ define amdgpu_ps void @insertelement_s_v2i8_v_s(ptr addrspace(4) inreg %ptr, i8
; GFX10-NEXT: global_store_short v[2:3], v0, off
; GFX10-NEXT: s_endpgm
;
-; GFX11-LABEL: insertelement_s_v2i8_v_s:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_mov_b32 m0, s4
-; GFX11-NEXT: global_load_u16 v1, v1, s[2:3]
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_readfirstlane_b32 s0, v1
-; GFX11-NEXT: s_lshr_b32 s1, s0, 8
-; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
-; GFX11-NEXT: v_movreld_b32_e32 v1, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v2
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; 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_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 m0, s4
+; GFX11-TRUE16-NEXT: global_load_u16 v1, v1, s[2:3]
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v1
+; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_movreld_b32_e32 v1, v0
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; 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: 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_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 m0, s4
+; GFX11-FAKE16-NEXT: global_load_u16 v1, v1, s[2:3]
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
+; GFX11-FAKE16-NEXT: v_movreld_b32_e32 v1, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xff, v2
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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
@@ -402,30 +463,56 @@ 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_readfirstlane_b32 s0, v1
-; GFX11-NEXT: s_lshr_b32 s1, s0, 8
-; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v1, s1
-; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: v_cndmask_b32_e64 v0, v2, s4, vcc_lo
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; 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_readfirstlane_b32 s0, v1
+; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, s0
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s1
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v2, v2, s4, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
+; 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_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_readfirstlane_b32 s0, v1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v1, s1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v2, s4, vcc_lo
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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
@@ -519,28 +606,52 @@ 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_mov_b32_e32 v2, 0
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
-; GFX11-NEXT: global_load_u16 v2, v2, s[2:3]
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_readfirstlane_b32 s0, v2
-; GFX11-NEXT: s_lshr_b32 s1, s0, 8
-; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_cndmask_b32_e32 v2, s1, v0, vcc_lo
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v2
-; GFX11-NEXT: v_cndmask_b32_e32 v0, s0, v0, vcc_lo
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; 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_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
+; GFX11-TRUE16-NEXT: global_load_u16 v2, v2, s[2:3]
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v2
+; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, s1, v0, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, s0, v0, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; 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_mov_b32_e32 v2, 0
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
+; GFX11-FAKE16-NEXT: global_load_u16 v2, v2, s[2:3]
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v2
+; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, s1, v0, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v2
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, s0, v0, vcc_lo
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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
@@ -623,26 +734,47 @@ 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_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; 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: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; 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 v2, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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
@@ -715,23 +847,41 @@ define amdgpu_ps void @insertelement_v_v2i8_v_s(ptr addrspace(1) %ptr, i8 %val,
; GFX10-NEXT: global_store_short v[1:2], v0, 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: s_mov_b32 m0, s2
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_lshrrev_b32_e32 v1, 8, v0
-; GFX11-NEXT: v_movreld_b32_e32 v0, v2
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v0
-; GFX11-NEXT: v_mov_b32_e32 v0, 0
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; 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 v3, v[0:1], off
+; GFX11-TRUE16-NEXT: s_mov_b32 m0, s2
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-TRUE16-NEXT: v_lshrrev_b16 v4.l, 8, v3.l
+; GFX11-TRUE16-NEXT: v_movreld_b32_e32 v3, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v4.l
+; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v3.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; 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: s_mov_b32 m0, s2
+; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
+; GFX11-FAKE16-NEXT: v_movreld_b32_e32 v0, v2
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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
@@ -811,25 +961,45 @@ 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_dual_mov_b32 v0, 0 :: v_dual_and_b32 v1, 0xff, v1
-; GFX11-NEXT: v_lshlrev_b16 v3, 8, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, 0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; 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_1)
+; 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.h, 0xff, v2.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
+; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; 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: 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_dual_mov_b32 v0, 0 :: v_dual_and_b32 v1, 0xff, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; 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/legalize-load-constant.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
index 8d3bd6d1704ee..0944642864770 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir
@@ -3,7 +3,8 @@
# 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=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 +38,15 @@ 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)
+ ;
+ ; 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 +85,15 @@ 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)
+ ;
+ ; 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 +126,13 @@ 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)
+ ;
+ ; 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 +165,13 @@ 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)
+ ;
+ ; 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 +204,21 @@ 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)
+ ;
+ ; 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 +251,21 @@ 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)
+ ;
+ ; 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 +316,19 @@ 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)
+ ;
+ ; 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 +361,13 @@ 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)
+ ;
+ ; 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 +417,19 @@ 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)
+ ;
+ ; 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 +509,29 @@ 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)
+ ;
+ ; 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 +563,13 @@ 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)
+ ;
+ ; 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 +602,13 @@ 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)
+ ;
+ ; 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 +659,19 @@ 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)
+ ;
+ ; 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 +740,25 @@ 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)
+ ;
+ ; 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 +797,15 @@ 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)
+ ;
+ ; 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 +838,13 @@ 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)
+ ;
+ ; 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 +876,13 @@ 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)
+ ;
+ ; 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 +968,31 @@ 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)
+ ;
+ ; 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 +1132,49 @@ 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)
+ ;
+ ; 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 +1209,14 @@ 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)
+ ;
+ ; 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 +1251,14 @@ 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)
+ ;
+ ; 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 +1293,14 @@ 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)
+ ;
+ ; 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 +1398,35 @@ 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)
+ ;
+ ; 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 +1602,61 @@ 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)
+ ;
+ ; 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 +1706,19 @@ 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)
+ ;
+ ; 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
@@ -1403,6 +1780,23 @@ body: |
; GFX9-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s224)
; GFX9-NEXT: [[MV:%[0-9]+]]:_(s256) = G_MERGE_VALUES [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[DEF]](s32)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[MV]](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]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s224)
+ ; GFX11-NEXT: [[MV:%[0-9]+]]:_(s256) = G_MERGE_VALUES [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[DEF]](s32)
+ ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[MV]](s256)
%0:_(p4) = COPY $vgpr0_vgpr1
%1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 4)
%2:_(s256) = G_IMPLICIT_DEF
@@ -1440,6 +1834,14 @@ 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)
+ ;
+ ; 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
@@ -1474,6 +1876,14 @@ 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)
+ ;
+ ; 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
@@ -1694,6 +2104,76 @@ 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)
+ ;
+ ; 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
@@ -1728,6 +2208,14 @@ 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)
+ ;
+ ; 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
@@ -1759,6 +2247,13 @@ 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)
+ ;
+ ; 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
@@ -1790,6 +2285,13 @@ 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)
+ ;
+ ; 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
@@ -1932,6 +2434,50 @@ 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)
+ ;
+ ; 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
@@ -1963,6 +2509,13 @@ 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)
+ ;
+ ; 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
@@ -1994,6 +2547,13 @@ 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)
+ ;
+ ; 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
@@ -2025,6 +2585,13 @@ 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)
+ ;
+ ; 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
@@ -2113,6 +2680,32 @@ 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)
+ ;
+ ; 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
@@ -2255,6 +2848,50 @@ 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)
+ ;
+ ; 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
@@ -2286,6 +2923,13 @@ 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)
+ ;
+ ; 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
@@ -2338,6 +2982,20 @@ 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)
+ ;
+ ; 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
@@ -2420,6 +3078,30 @@ 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)
+ ;
+ ; 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
@@ -2451,6 +3133,21 @@ 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)
+ ;
+ ; 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
@@ -2484,6 +3181,21 @@ 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)
+ ;
+ ; 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
@@ -2535,6 +3247,34 @@ 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)
+ ;
+ ; 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
@@ -2639,6 +3379,63 @@ 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)
+ ;
+ ; 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
@@ -2773,6 +3570,83 @@ 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)
+ ;
+ ; 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
@@ -2806,6 +3680,13 @@ 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)
+ ;
+ ; 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
@@ -2856,6 +3737,19 @@ 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)
+ ;
+ ; 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
@@ -2936,6 +3830,29 @@ 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)
+ ;
+ ; 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
@@ -2968,6 +3885,13 @@ 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>)
+ ;
+ ; 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
@@ -3099,6 +4023,62 @@ 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>)
+ ;
+ ; 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
@@ -3313,6 +4293,106 @@ 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>)
+ ;
+ ; 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
@@ -3345,6 +4425,13 @@ 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>)
+ ;
+ ; 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
@@ -3402,6 +4489,30 @@ 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>)
+ ;
+ ; 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
@@ -3489,6 +4600,29 @@ 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>)
+ ;
+ ; 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
@@ -3538,6 +4672,30 @@ 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>)
+ ;
+ ; 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
@@ -3629,6 +4787,45 @@ 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>)
+ ;
+ ; 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
@@ -3720,6 +4917,45 @@ 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>)
+ ;
+ ; 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
@@ -3853,6 +5089,41 @@ 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>)
+ ;
+ ; 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
@@ -3886,6 +5157,13 @@ 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>)
+ ;
+ ; 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
@@ -3917,6 +5195,13 @@ 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>)
+ ;
+ ; 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
@@ -4008,6 +5293,48 @@ 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>)
+ ;
+ ; 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
@@ -4153,6 +5480,47 @@ 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>)
+ ;
+ ; 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
@@ -4187,6 +5555,14 @@ 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>)
+ ;
+ ; 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
@@ -4218,6 +5594,13 @@ 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>)
+ ;
+ ; 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
@@ -4249,6 +5632,13 @@ 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>)
+ ;
+ ; 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
@@ -4322,6 +5712,27 @@ 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>)
+ ;
+ ; 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
@@ -4449,6 +5860,45 @@ 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>)
+ ;
+ ; 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
@@ -4480,6 +5930,13 @@ 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>)
+ ;
+ ; 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
@@ -4513,6 +5970,13 @@ 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>)
+ ;
+ ; 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
@@ -4544,6 +6008,13 @@ 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>)
+ ;
+ ; 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
@@ -4575,6 +6046,13 @@ 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>)
+ ;
+ ; 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
@@ -4606,6 +6084,13 @@ 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>)
+ ;
+ ; 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
@@ -4637,6 +6122,13 @@ 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>)
+ ;
+ ; 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
@@ -4668,6 +6160,13 @@ 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>)
+ ;
+ ; 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
@@ -4699,6 +6198,13 @@ 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>)
+ ;
+ ; 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
@@ -4730,6 +6236,13 @@ 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>)
+ ;
+ ; 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
@@ -4761,6 +6274,13 @@ 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>)
+ ;
+ ; 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
@@ -4792,6 +6312,13 @@ 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>)
+ ;
+ ; 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
@@ -4934,6 +6461,50 @@ 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>)
+ ;
+ ; 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
@@ -5178,6 +6749,84 @@ 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>)
+ ;
+ ; 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
@@ -5209,6 +6858,13 @@ 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>)
+ ;
+ ; 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
@@ -5260,6 +6916,19 @@ 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>)
+ ;
+ ; 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
@@ -5611,6 +7280,119 @@ 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>)
+ ;
+ ; 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
@@ -5644,6 +7426,13 @@ 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>)
+ ;
+ ; 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
@@ -5675,6 +7464,13 @@ 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>)
+ ;
+ ; 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
@@ -6123,6 +7919,152 @@ 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>)
+ ;
+ ; 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
@@ -6157,6 +8099,14 @@ 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>)
+ ;
+ ; 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
@@ -6191,6 +8141,14 @@ 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>)
+ ;
+ ; 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
@@ -6225,6 +8183,14 @@ 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>)
+ ;
+ ; 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
@@ -6259,6 +8225,14 @@ 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>)
+ ;
+ ; 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
@@ -6479,6 +8453,76 @@ 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>)
+ ;
+ ; 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
@@ -6513,6 +8557,14 @@ 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>)
+ ;
+ ; 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
@@ -6547,6 +8599,14 @@ 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>)
+ ;
+ ; 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
@@ -6677,6 +8737,46 @@ 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>)
+ ;
+ ; 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
@@ -6708,6 +8808,13 @@ 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)
+ ;
+ ; 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
@@ -6739,6 +8846,13 @@ 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)
+ ;
+ ; 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
@@ -6774,6 +8888,14 @@ 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)
+ ;
+ ; 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
@@ -6808,6 +8930,22 @@ 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)
+ ;
+ ; 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
@@ -6842,6 +8980,14 @@ 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)
+ ;
+ ; 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
@@ -6885,6 +9031,17 @@ 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)
+ ;
+ ; 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
@@ -6919,6 +9076,22 @@ 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)
+ ;
+ ; 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
@@ -6953,6 +9126,14 @@ 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)
+ ;
+ ; 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
@@ -7080,6 +9261,45 @@ 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>)
+ ;
+ ; 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
@@ -7153,6 +9373,27 @@ 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>)
+ ;
+ ; 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
@@ -7184,6 +9425,13 @@ 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>)
+ ;
+ ; 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
@@ -7215,6 +9463,13 @@ 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>)
+ ;
+ ; 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
@@ -7246,6 +9501,13 @@ 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>)
+ ;
+ ; 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
@@ -7565,6 +9827,109 @@ 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)
+ ;
+ ; 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
@@ -7737,6 +10102,59 @@ 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)
+ ;
+ ; 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
@@ -7795,6 +10213,21 @@ 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)
+ ;
+ ; 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
@@ -7853,6 +10286,21 @@ 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)
+ ;
+ ; 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
@@ -7890,6 +10338,14 @@ 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)
+ ;
+ ; 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
@@ -7924,6 +10380,14 @@ 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>)
+ ;
+ ; 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 4da6bc03708bc..32e617c6d963f 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir
@@ -3,15 +3,19 @@
# 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=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=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
+# 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,UNALIGNEDGFX11PLUS-TRUE16 %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,UNALIGNED_GFX11PLUS-FAKE16 %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,UNALIGNED_GFX12-TRUE16 %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,UNALIGNED_GFX12-FAKE16 %s
---
name: test_load_flat_s1_align1
@@ -340,19 +344,35 @@ 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)
+ ; 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
@@ -361,19 +381,35 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align4
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_s16_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align4
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_s16_align4
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_s16_align4
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; 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_flat_s16_align4
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 4, addrspace 0)
%2:_(s32) = G_ANYEXT %1
@@ -407,19 +443,35 @@ 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)
+ ; 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
@@ -428,19 +480,35 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_s16_align2
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_s16_align2
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_s16_align2
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_s16_align2
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; 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_flat_s16_align2
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s16) = G_LOAD %0 :: (load (s16), align 2, addrspace 0)
%2:_(s32) = G_ANYEXT %1
@@ -486,19 +554,35 @@ 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)
+ ; 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
@@ -4608,19 +4692,35 @@ 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)
+ ; 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
@@ -4629,19 +4729,35 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align4
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v2s8_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align4
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v2s8_align4
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v2s8_align4
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; 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_flat_v2s8_align4
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 4, addrspace 0)
%2:_(s16) = G_BITCAST %1
@@ -4676,19 +4792,35 @@ 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)
+ ; 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
@@ -4697,19 +4829,35 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
;
- ; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align2
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[LOAD]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v2s8_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align2
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v2s8_align2
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v2s8_align2
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; 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_flat_v2s8_align2
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-FAKE16-NEXT: $vgpr0 = COPY [[LOAD]](s32)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 2, addrspace 0)
%2:_(s16) = G_BITCAST %1
@@ -4756,19 +4904,35 @@ 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)
+ ; 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
@@ -4783,31 +4947,61 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
;
- ; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v2s8_align1
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[OR]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v2s8_align1
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX12-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX12-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s8_align1
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[OR]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v2s8_align1
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[OR]](s32)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v2s8_align1
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[OR]](s32)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s16)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; UNALIGNED_GFX12-FAKE16-LABEL: name: test_load_flat_v2s8_align1
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX12-FAKE16-NEXT: $vgpr0 = COPY [[OR]](s32)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(<2 x s8>) = G_LOAD %0 :: (load (<2 x s8>), align 1, addrspace 0)
%2:_(s16) = G_BITCAST %1
@@ -4913,65 +5107,119 @@ body: |
; GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; GFX9PLUS-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
@@ -5003,65 +5251,119 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
;
- ; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s8_align4
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; UNALIGNED_GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX11PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX11PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[OR2]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v3s8_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
- ; 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)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s8_align4
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s8_align4
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[OR2]](s32)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v3s8_align4
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; 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_flat_v3s8_align4
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32))
+ ; 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:_(p0) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 4, addrspace 0)
%2:_(s24) = G_BITCAST %1
@@ -5192,75 +5494,139 @@ body: |
; GFX9PLUS-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
; GFX9PLUS-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
@@ -5302,85 +5668,159 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
; UNALIGNED_GFX9PLUS-NEXT: $vgpr0 = COPY [[OR4]](s32)
;
- ; UNALIGNED_GFX11PLUS-LABEL: name: test_load_flat_v3s8_align1
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
- ; UNALIGNED_GFX11PLUS-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[OR4]](s32)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v3s8_align1
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX12-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
- ; 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]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
- ; 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)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s8_align1
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s8_align1
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[OR4]](s32)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v3s8_align1
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; 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]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; 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_flat_v3s8_align1
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
+ ; 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]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s8) from unknown-address + 2)
+ ; 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:_(p0) = COPY $vgpr0_vgpr1
%1:_(<3 x s8>) = G_LOAD %0 :: (load (<3 x s8>), align 1, addrspace 0)
%2:_(s24) = G_BITCAST %1
@@ -6137,31 +6577,53 @@ 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_GFX11PLUS-LABEL: name: test_load_flat_v2s16_align2
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v2s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; 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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v2s16_align2
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v2s16_align2
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v2s16_align2
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; 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_flat_v2s16_align2
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; 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:_(p0) = COPY $vgpr0_vgpr1
%1:_(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 2, addrspace 0)
$vgpr0 = COPY %1
@@ -6392,27 +6854,53 @@ 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>)
;
- ; 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>)
- ;
- ; 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-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>)
+ ;
+ ; 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
@@ -6425,27 +6913,53 @@ 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_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align8
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; UNALIGNED_GFX11PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
- ; UNALIGNED_GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_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>)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v3s16_align8
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
- ; 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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align8
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; UNALIGNEDGFX11PLUS-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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNEDGFX11PLUS-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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s16_align8
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_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>)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v3s16_align8
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; 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_flat_v3s16_align8
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
+ ; 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:_(p0) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 8, addrspace 0)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -6534,47 +7048,83 @@ 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>)
+ ; 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
@@ -6597,47 +7147,83 @@ 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_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align4
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_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>)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v3s16_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
- ; 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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align4
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNEDGFX11PLUS-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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s16_align4
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_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>)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v3s16_align4
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; 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_flat_v3s16_align4
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4, align 4)
+ ; 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:_(p0) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 4, addrspace 0)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -6730,47 +7316,83 @@ 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>)
+ ; 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
@@ -6793,47 +7415,83 @@ 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_GFX11PLUS-LABEL: name: test_load_flat_v3s16_align2
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
- ; UNALIGNED_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>)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v3s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; 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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v3s16_align2
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNEDGFX11PLUS-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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v3s16_align2
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; UNALIGNED_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>)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v3s16_align2
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; 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_flat_v3s16_align2
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; 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:_(p0) = COPY $vgpr0_vgpr1
%1:_(<3 x s16>) = G_LOAD %0 :: (load (<3 x s16>), align 2, addrspace 0)
%2:_(<3 x s16>) = G_IMPLICIT_DEF
@@ -6954,47 +7612,83 @@ 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>)
+ ; 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
@@ -7361,51 +8055,89 @@ 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_GFX11PLUS-LABEL: name: test_load_flat_v4s16_align2
- ; UNALIGNED_GFX11PLUS: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; UNALIGNED_GFX11PLUS-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
- ; UNALIGNED_GFX11PLUS-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
- ; UNALIGNED_GFX11PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_load_flat_v4s16_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
- ; UNALIGNED_GFX12-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
- ; UNALIGNED_GFX12-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
- ; UNALIGNED_GFX12-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
- ; UNALIGNED_GFX12-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
- ; UNALIGNED_GFX12-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
- ; 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>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_load_flat_v4s16_align2
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_load_flat_v4s16_align2
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_load_flat_v4s16_align2
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; 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_flat_v4s16_align2
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16))
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s16) from unknown-address + 2)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load (s16) from unknown-address + 4)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p0) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p0) :: (load (s16) from unknown-address + 6)
+ ; 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:_(p0) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 2, addrspace 0)
$vgpr0_vgpr1 = COPY %1
@@ -12981,21 +13713,37 @@ 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)
+ ; 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
@@ -13005,21 +13753,37 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-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: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX11PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_ext_load_flat_s64_from_2_align4
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align4
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; 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_flat_s64_from_2_align4
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 0)
$vgpr0_vgpr1 = COPY %1
@@ -13227,21 +13991,37 @@ 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)
+ ; 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
@@ -13251,21 +14031,37 @@ body: |
; UNALIGNED_GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
; UNALIGNED_GFX9PLUS-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: {{ $}}
- ; UNALIGNED_GFX11PLUS-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX11PLUS-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX11PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX11PLUS-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
- ;
- ; UNALIGNED_GFX12-LABEL: name: test_ext_load_flat_s64_from_2_align2
- ; UNALIGNED_GFX12: liveins: $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: {{ $}}
- ; UNALIGNED_GFX12-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
- ; UNALIGNED_GFX12-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
- ; UNALIGNED_GFX12-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
- ; UNALIGNED_GFX12-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ; UNALIGNEDGFX11PLUS-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; UNALIGNEDGFX11PLUS-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: {{ $}}
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; UNALIGNEDGFX11PLUS-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX11PLUS-FAKE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; UNALIGNED_GFX11PLUS-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX11PLUS-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; UNALIGNED_GFX12-TRUE16-LABEL: name: test_ext_load_flat_s64_from_2_align2
+ ; UNALIGNED_GFX12-TRUE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; 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_flat_s64_from_2_align2
+ ; UNALIGNED_GFX12-FAKE16: liveins: $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: {{ $}}
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; UNALIGNED_GFX12-FAKE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s64) = G_LOAD %0 :: (load (s16), align 4, addrspace 0)
$vgpr0_vgpr1 = COPY %1
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
index ea6832d417b84..92c61e05f72de 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir
@@ -7,17 +7,21 @@
# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=bonaire -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=CI-MESA %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-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 -pass-remarks-missed='gisel.*' 2> %t.err %s -o - | FileCheck -check-prefix=GFX9-HSA %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 -pass-remarks-missed='gisel.*' 2> %t.err %s -o - | FileCheck -check-prefixes=GFX9-HSA %s
# RUN: FileCheck -check-prefix=ERR %s < %t.err
# 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=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,GFX12-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,GFX12-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 +90,42 @@ 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)
+ ;
+ ; 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 +191,42 @@ 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)
+ ;
+ ; 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 +280,34 @@ 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)
+ ;
+ ; 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 +361,34 @@ 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)
+ ;
+ ; 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 +442,66 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_s16_align4
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_s16_align4
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-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 +555,66 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_s16_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_s16_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-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 +692,62 @@ 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)
+ ;
+ ; 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 +801,34 @@ 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)
+ ;
+ ; 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 +905,46 @@ 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)
+ ;
+ ; 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 +1061,66 @@ 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)
+ ;
+ ; 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 +1173,34 @@ 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)
+ ;
+ ; 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 +1254,34 @@ 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)
+ ;
+ ; 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 +1371,58 @@ 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)
+ ;
+ ; 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 +1536,70 @@ 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)
+ ;
+ ; 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 +1665,42 @@ 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)
+ ;
+ ; 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 +1754,34 @@ 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)
+ ;
+ ; 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 +1834,34 @@ 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)
+ ;
+ ; 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 +1986,70 @@ 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)
+ ;
+ ; 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 +2246,106 @@ 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)
+ ;
+ ; 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 +2406,38 @@ 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)
+ ;
+ ; 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 +2501,38 @@ 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)
+ ;
+ ; 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 +2596,38 @@ 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)
+ ;
+ ; 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 +2770,80 @@ 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)
+ ;
+ ; 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 +3090,132 @@ 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)
+ ;
+ ; 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 +3304,58 @@ 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)
+ ;
+ ; 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
@@ -2259,6 +3471,74 @@ body: |
; GFX9-MESA-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s224)
; GFX9-MESA-NEXT: [[MV:%[0-9]+]]:_(s256) = G_MERGE_VALUES [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[DEF]](s32)
; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[MV]](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]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-HSA-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s224)
+ ; GFX11-HSA-NEXT: [[MV:%[0-9]+]]:_(s256) = G_MERGE_VALUES [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[DEF]](s32)
+ ; GFX11-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[MV]](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]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX11-MESA-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s224)
+ ; GFX11-MESA-NEXT: [[MV:%[0-9]+]]:_(s256) = G_MERGE_VALUES [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[DEF]](s32)
+ ; GFX11-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[MV]](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]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-HSA-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s224)
+ ; GFX12-HSA-NEXT: [[MV:%[0-9]+]]:_(s256) = G_MERGE_VALUES [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[DEF]](s32)
+ ; GFX12-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[MV]](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]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s224)
+ ; GFX12-MESA-NEXT: [[MV:%[0-9]+]]:_(s256) = G_MERGE_VALUES [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[DEF]](s32)
+ ; GFX12-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[MV]](s256)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 1)
%2:_(s256) = G_IMPLICIT_DEF
@@ -2320,6 +3600,38 @@ 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)
+ ;
+ ; 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
@@ -2378,6 +3690,38 @@ 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)
+ ;
+ ; 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
@@ -2684,6 +4028,162 @@ 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)
+ ;
+ ; 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
@@ -2742,6 +4242,38 @@ 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)
+ ;
+ ; 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
@@ -2794,6 +4326,34 @@ 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)
+ ;
+ ; 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
@@ -2846,6 +4406,34 @@ 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)
+ ;
+ ; 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
@@ -3046,6 +4634,108 @@ 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)
+ ;
+ ; 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
@@ -3098,6 +4788,34 @@ 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)
+ ;
+ ; 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
@@ -3150,6 +4868,34 @@ 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)
+ ;
+ ; 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
@@ -3202,6 +4948,34 @@ 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)
+ ;
+ ; 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
@@ -3330,6 +5104,72 @@ 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)
+ ;
+ ; 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
@@ -3530,6 +5370,108 @@ 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)
+ ;
+ ; 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
@@ -3582,6 +5524,34 @@ 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)
+ ;
+ ; 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
@@ -3662,6 +5632,48 @@ 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)
+ ;
+ ; 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
@@ -3782,6 +5794,68 @@ 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)
+ ;
+ ; 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
@@ -3834,6 +5908,66 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align4
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-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
@@ -3888,6 +6022,66 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-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
@@ -3966,6 +6160,92 @@ 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)
+ ;
+ ; 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-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
+ ; 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-TRUE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-TRUE16-NEXT: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[OR]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v2s8_align1
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-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
@@ -4164,6 +6444,234 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-MESA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-MESA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-MESA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR]](s16), [[OR1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v3s8_align4
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-MESA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C2]]
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C2]]
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C2]]
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C2]]
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
+ ; GFX12-MESA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
+ ; GFX12-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
@@ -4413,6 +6921,294 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-TRUE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-TRUE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-MESA-TRUE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-MESA-TRUE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX12-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-MESA-TRUE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX12-MESA-TRUE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-MESA-TRUE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX12-MESA-TRUE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX12-MESA-TRUE16-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[OR2]](s16), [[OR3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[MV]](s32)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v3s8_align1
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]]
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s8) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[SHL1]], [[OR]]
+ ; GFX12-MESA-FAKE16-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR1]], [[C3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[OR1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
+ ; GFX12-MESA-FAKE16-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C4]]
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C4]]
+ ; GFX12-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C5]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]]
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C4]]
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C4]]
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C5]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL3]]
+ ; GFX12-MESA-FAKE16-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
+ ; GFX12-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
@@ -4527,6 +7323,90 @@ 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>)
+ ;
+ ; 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
@@ -4659,6 +7539,100 @@ 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>)
+ ;
+ ; 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
@@ -4827,6 +7801,118 @@ 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>)
+ ;
+ ; 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
@@ -4879,6 +7965,34 @@ 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>)
+ ;
+ ; 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
@@ -4932,6 +8046,34 @@ 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>)
+ ;
+ ; 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
@@ -4985,6 +8127,34 @@ 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>)
+ ;
+ ; 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
@@ -5039,6 +8209,34 @@ 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>)
+ ;
+ ; 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
@@ -5127,6 +8325,68 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v2s16_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v2s16_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-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
@@ -5255,6 +8515,66 @@ 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>)
+ ;
+ ; 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
@@ -5343,6 +8663,102 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; 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-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-TRUE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-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>)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-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>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align8
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-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>)
+ ; GFX12-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
@@ -5513,6 +8929,162 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-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>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align4
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-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>)
+ ; GFX12-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
@@ -5683,6 +9255,162 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-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>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v3s16_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[DEF]](s16), [[DEF]](s16)
+ ; GFX12-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>)
+ ; GFX12-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
@@ -5909,6 +9637,154 @@ 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>)
+ ;
+ ; 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
@@ -5963,6 +9839,34 @@ 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>)
+ ;
+ ; 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
@@ -6015,6 +9919,34 @@ 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>)
+ ;
+ ; 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
@@ -6149,6 +10081,104 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v4s16_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v4s16_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<2 x s16>), [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-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
@@ -6355,6 +10385,102 @@ 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>)
+ ;
+ ; 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
@@ -6435,6 +10561,118 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align16
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-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
@@ -6645,6 +10883,226 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align8
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-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
@@ -6855,6 +11313,226 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align4
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-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
@@ -7083,6 +11761,226 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v5s16_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-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
@@ -7399,6 +12297,220 @@ 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>)
+ ;
+ ; 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
@@ -7465,6 +12577,38 @@ 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>)
+ ;
+ ; 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
@@ -7528,6 +12672,38 @@ 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>)
+ ;
+ ; 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
@@ -7591,6 +12767,38 @@ 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>)
+ ;
+ ; 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
@@ -7733,6 +12941,80 @@ 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>)
+ ;
+ ; 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
@@ -7979,6 +13261,132 @@ 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>)
+ ;
+ ; 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
@@ -8067,6 +13475,134 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[UV8]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[UV13]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[UV18]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align16
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-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>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[UV4]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[UV9]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr2 = COPY [[UV14]](<2 x s16>)
+ ; GFX12-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
@@ -8364,6 +13900,298 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align8
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-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
@@ -8661,6 +14489,298 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align4
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-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
@@ -8958,6 +15078,298 @@ 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>)
+ ;
+ ; 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>)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD1:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD2:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD4:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD5:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-TRUE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD6:%[0-9]+]]:_(s16) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD]](s16), [[LOAD1]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD2]](s16), [[LOAD3]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD4]](s16), [[LOAD5]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-TRUE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[LOAD6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr3 = COPY [[BUILD_VECTOR3]](<2 x s16>)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_load_global_v7s16_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C1]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C2]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD3]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C3]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD4]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 10
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C4]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD5]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
+ ; GFX12-MESA-FAKE16-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = nuw inbounds G_PTR_ADD [[COPY]], [[C5]](s64)
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD6]](s32)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+ ; GFX12-MESA-FAKE16-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[DEF]](s16)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr1 = COPY [[BUILD_VECTOR1]](<2 x s16>)
+ ; GFX12-MESA-FAKE16-NEXT: $vgpr2 = COPY [[BUILD_VECTOR2]](<2 x s16>)
+ ; GFX12-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
@@ -9375,6 +15787,292 @@ 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>)
+ ;
+ ; 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
@@ -9440,6 +16138,38 @@ 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>)
+ ;
+ ; 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
@@ -9498,6 +16228,38 @@ 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>)
+ ;
+ ; 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
@@ -9550,6 +16312,34 @@ 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>)
+ ;
+ ; 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
@@ -9602,6 +16392,34 @@ 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>)
+ ;
+ ; 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
@@ -9710,6 +16528,62 @@ 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>)
+ ;
+ ; 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
@@ -9890,6 +16764,98 @@ 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>)
+ ;
+ ; 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
@@ -9944,6 +16910,34 @@ 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>)
+ ;
+ ; 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
@@ -10001,6 +16995,34 @@ 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>)
+ ;
+ ; 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
@@ -10053,6 +17075,34 @@ 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>)
+ ;
+ ; 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
@@ -10105,6 +17155,34 @@ 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>)
+ ;
+ ; 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
@@ -10157,6 +17235,34 @@ 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>)
+ ;
+ ; 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
@@ -10209,6 +17315,34 @@ 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>)
+ ;
+ ; 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
@@ -10261,6 +17395,34 @@ 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>)
+ ;
+ ; 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
@@ -10313,6 +17475,34 @@ 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>)
+ ;
+ ; 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
@@ -10365,6 +17555,34 @@ 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>)
+ ;
+ ; 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
@@ -10417,6 +17635,34 @@ 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>)
+ ;
+ ; 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
@@ -10617,6 +17863,108 @@ 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>)
+ ;
+ ; 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
@@ -10953,6 +18301,176 @@ 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>)
+ ;
+ ; 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
@@ -11011,6 +18529,38 @@ 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>)
+ ;
+ ; 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
@@ -11063,6 +18613,34 @@ 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>)
+ ;
+ ; 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
@@ -11153,6 +18731,58 @@ 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>)
+ ;
+ ; 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
@@ -11643,6 +19273,258 @@ 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>)
+ ;
+ ; 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
@@ -11697,6 +19579,34 @@ 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>)
+ ;
+ ; 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
@@ -11749,6 +19659,34 @@ 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>)
+ ;
+ ; 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
@@ -12357,6 +20295,312 @@ 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>)
+ ;
+ ; 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
@@ -12415,6 +20659,38 @@ 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>)
+ ;
+ ; 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
@@ -12473,6 +20749,38 @@ 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>)
+ ;
+ ; 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
@@ -12531,6 +20839,38 @@ 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>)
+ ;
+ ; 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
@@ -12589,6 +20929,38 @@ 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>)
+ ;
+ ; 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
@@ -12895,6 +21267,162 @@ 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>)
+ ;
+ ; 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
@@ -12953,6 +21481,38 @@ 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>)
+ ;
+ ; 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
@@ -13011,6 +21571,38 @@ 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>)
+ ;
+ ; 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
@@ -13069,6 +21661,38 @@ 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>)
+ ;
+ ; 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
@@ -13255,6 +21879,102 @@ 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>)
+ ;
+ ; 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
@@ -13307,6 +22027,34 @@ 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)
+ ;
+ ; 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
@@ -13359,6 +22107,34 @@ 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)
+ ;
+ ; 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
@@ -13471,6 +22247,70 @@ 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)
+ ;
+ ; 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
@@ -13558,6 +22398,58 @@ 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)
+ ;
+ ; 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
@@ -13610,6 +22502,34 @@ 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)
+ ;
+ ; 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
@@ -13669,6 +22589,38 @@ 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)
+ ;
+ ; 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
@@ -13727,6 +22679,70 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align4
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-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
@@ -13785,6 +22801,38 @@ 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)
+ ;
+ ; 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
@@ -13861,6 +22909,50 @@ 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)
+ ;
+ ; 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
@@ -13919,6 +23011,70 @@ 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)
+ ;
+ ; 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)
+ ;
+ ; GFX12-MESA-TRUE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX12-MESA-TRUE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: {{ $}}
+ ; GFX12-MESA-TRUE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-TRUE16-NEXT: [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-TRUE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s16)
+ ; GFX12-MESA-TRUE16-NEXT: $vgpr0_vgpr1 = COPY [[ANYEXT]](s64)
+ ;
+ ; GFX12-MESA-FAKE16-LABEL: name: test_ext_load_global_s64_from_2_align2
+ ; GFX12-MESA-FAKE16: liveins: $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: {{ $}}
+ ; GFX12-MESA-FAKE16-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
+ ; GFX12-MESA-FAKE16-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
+ ; GFX12-MESA-FAKE16-NEXT: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[LOAD]](s32)
+ ; GFX12-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
@@ -13977,6 +23133,38 @@ 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)
+ ;
+ ; 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
@@ -14029,6 +23217,34 @@ 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>)
+ ;
+ ; 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
@@ -14081,6 +23297,34 @@ 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>)
+ ;
+ ; 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
@@ -14133,6 +23377,34 @@ 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>)
+ ;
+ ; 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
@@ -14185,6 +23457,34 @@ 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>)
+ ;
+ ; 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
@@ -14237,6 +23537,34 @@ 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>)
+ ;
+ ; 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
@@ -14689,6 +24017,242 @@ 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)
+ ;
+ ; 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
@@ -14944,6 +24508,142 @@ 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)
+ ;
+ ; 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
@@ -15056,6 +24756,66 @@ 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)
+ ;
+ ; 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
@@ -15166,6 +24926,66 @@ 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)
+ ;
+ ; 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
@@ -15701,6 +25521,482 @@ 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>)
+ ;
+ ; 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
@@ -15873,6 +26169,146 @@ 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>)
+ ;
+ ; 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
@@ -15926,6 +26362,34 @@ 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)
+ ;
+ ; 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..729218b239b9a 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 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode,+real-true16 -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode,-real-true16 -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode,+real-true16 -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-UNALIGNED,GFX11-UNALIGNED-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode,-real-true16 -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX11-UNALIGNED,GFX11-UNALIGNED-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode,+real-true16 -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -O0 -run-pass=legalizer -mattr=-unaligned-access-mode,-real-true16 -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12,GFX12-FAKE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode,+real-true16 -global-isel-abort=0 %s -o - | FileCheck -check-prefixes=GFX12-UNALIGNED,GFX12-UNALIGNED-TRUE16 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -O0 -run-pass=legalizer -mattr=+unaligned-access-mode,-real-true16 -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
@@ -7387,65 +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)
+ ; 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
@@ -7775,80 +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)
+ ; 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
@@ -7931,6 +9568,20 @@ 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 +9663,20 @@ 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 +10193,81 @@ 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 +10349,20 @@ 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 +10482,29 @@ 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 +10512,37 @@ 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 +10751,36 @@ 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
@@ -9066,27 +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-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-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-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>)
+ ;
+ ; 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>)
+ ;
+ ; 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
@@ -9300,47 +11200,161 @@ 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 +11687,118 @@ 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 +11880,20 @@ 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 +12028,20 @@ 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 +12244,47 @@ 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 +12292,55 @@ 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 +12699,54 @@ 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 +12827,20 @@ 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 +12921,20 @@ 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 +13117,34 @@ 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 +13453,52 @@ 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 +13908,66 @@ 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 +14100,27 @@ 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 +14209,20 @@ 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 +14331,20 @@ 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 +14493,30 @@ 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 +14805,48 @@ 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 +15366,81 @@ 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 +15573,28 @@ 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 +15799,40 @@ 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 +15945,24 @@ 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 +16552,90 @@ 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 +16780,32 @@ 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 +16940,28 @@ 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 +17120,32 @@ 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 +17236,22 @@ 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 +17332,20 @@ 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 +17426,20 @@ 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 +17531,22 @@ 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 +17622,69 @@ 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 +17775,22 @@ 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 +17911,28 @@ 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 +18008,69 @@ 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 +18161,22 @@ 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 +18257,20 @@ 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 +18351,20 @@ 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 +18445,20 @@ 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 +18539,20 @@ 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 +18633,20 @@ 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 +19428,123 @@ 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 +19986,74 @@ 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 +20307,48 @@ 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 +20572,43 @@ 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..dbd7ea73525f2 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
@@ -7788,65 +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
@@ -7908,65 +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
@@ -8174,75 +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
@@ -8324,85 +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
@@ -9347,31 +9733,53 @@ 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
@@ -9717,27 +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>)
- ;
- ; 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-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>)
+ ;
+ ; 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
@@ -9781,27 +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_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-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_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
@@ -9944,47 +10404,83 @@ 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 +10524,83 @@ 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 +10785,83 @@ 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 +11407,89 @@ 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 +17640,37 @@ 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 +17688,37 @@ 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 +17999,37 @@ 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 +18047,37 @@ 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..45cfd6f315f1a 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir
@@ -4,8 +4,10 @@
# 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=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 +54,26 @@ 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)
+ ;
+ ; 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 +121,24 @@ 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)
+ ;
+ ; 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 +178,20 @@ 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)
+ ;
+ ; 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 +245,36 @@ 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)
+ ;
+ ; 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 +314,36 @@ 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)
+ ;
+ ; 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 +383,36 @@ 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)
+ ;
+ ; 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 +476,32 @@ 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)
+ ;
+ ; 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 +565,32 @@ 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)
+ ;
+ ; 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 +668,32 @@ 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)
+ ;
+ ; 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 +741,24 @@ 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)
+ ;
+ ; 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 +857,20 @@ 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)
+ ;
+ ; 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 +921,20 @@ 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)
+ ;
+ ; 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 +973,20 @@ 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)
+ ;
+ ; 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 +1062,20 @@ 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)
+ ;
+ ; 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 +1128,20 @@ 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)
+ ;
+ ; 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 +1180,20 @@ 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)
+ ;
+ ; 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 +1309,66 @@ 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)
+ ;
+ ; 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 +1452,66 @@ 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)
+ ;
+ ; 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 +1654,20 @@ 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)
+ ;
+ ; 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 +1742,20 @@ 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)
+ ;
+ ; 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 +1794,20 @@ 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)
+ ;
+ ; 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 +1846,20 @@ 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)
+ ;
+ ; 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 +1898,20 @@ 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)
+ ;
+ ; 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 +2030,20 @@ 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)
+ ;
+ ; 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 +2120,20 @@ 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)
+ ;
+ ; 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 +2172,20 @@ 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)
+ ;
+ ; 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 +2224,20 @@ 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)
+ ;
+ ; 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 +2276,20 @@ 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)
+ ;
+ ; 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 +2408,20 @@ 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)
+ ;
+ ; 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 +2498,20 @@ 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)
+ ;
+ ; 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 +2550,20 @@ 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)
+ ;
+ ; 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 +2602,20 @@ 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)
+ ;
+ ; 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 +2654,20 @@ 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)
+ ;
+ ; 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 +2776,20 @@ 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)
+ ;
+ ; 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 +2856,20 @@ 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)
+ ;
+ ; 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 +2908,20 @@ 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)
+ ;
+ ; 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 +2960,20 @@ 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)
+ ;
+ ; 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 +3012,20 @@ 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)
+ ;
+ ; 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 +3138,22 @@ 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)
+ ;
+ ; 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 +3224,22 @@ 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)
+ ;
+ ; 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 +3282,22 @@ 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)
+ ;
+ ; 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 +3340,22 @@ 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)
+ ;
+ ; 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 +3398,22 @@ 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)
+ ;
+ ; 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 +3526,20 @@ 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)
+ ;
+ ; 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 +3608,20 @@ 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)
+ ;
+ ; 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 +3660,20 @@ 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)
+ ;
+ ; 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 +3712,20 @@ 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)
+ ;
+ ; 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 +3764,20 @@ 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)
+ ;
+ ; 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 +3919,20 @@ 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)
+ ;
+ ; 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 +4013,20 @@ 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)
+ ;
+ ; 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 +4070,20 @@ 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)
+ ;
+ ; 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 +4127,20 @@ 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)
+ ;
+ ; 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 +4184,20 @@ 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)
+ ;
+ ; 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 +4372,20 @@ 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)
+ ;
+ ; 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 +4480,20 @@ 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)
+ ;
+ ; 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 +4532,20 @@ 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)
+ ;
+ ; 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 +4584,20 @@ 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)
+ ;
+ ; 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 +4636,20 @@ 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)
+ ;
+ ; 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 +4842,20 @@ 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)
+ ;
+ ; 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 +4968,20 @@ 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)
+ ;
+ ; 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 +5020,20 @@ 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)
+ ;
+ ; 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 +5072,20 @@ 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)
+ ;
+ ; 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 +5124,20 @@ 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)
+ ;
+ ; 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 +5316,22 @@ 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)
+ ;
+ ; 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 +5430,22 @@ 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)
+ ;
+ ; 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 +5488,22 @@ 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)
+ ;
+ ; 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 +5546,22 @@ 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)
+ ;
+ ; 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 +5604,22 @@ 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)
+ ;
+ ; 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 +5798,22 @@ 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)
+ ;
+ ; 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 +5912,22 @@ 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)
+ ;
+ ; 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 +5970,22 @@ 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)
+ ;
+ ; 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 +6028,22 @@ 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)
+ ;
+ ; 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 +6086,22 @@ 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)
+ ;
+ ; 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 +6247,22 @@ 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)
+ ;
+ ; 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 +6347,22 @@ 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)
+ ;
+ ; 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 +6410,22 @@ 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)
+ ;
+ ; 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 +6473,22 @@ 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)
+ ;
+ ; 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 +6536,22 @@ 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)
+ ;
+ ; 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 +6730,22 @@ 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)
+ ;
+ ; 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 +6844,22 @@ 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)
+ ;
+ ; 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 +6902,22 @@ 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)
+ ;
+ ; 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 +6960,22 @@ 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)
+ ;
+ ; 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 +7018,22 @@ 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)
+ ;
+ ; 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 +7251,30 @@ 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)
+ ;
+ ; 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 +7393,30 @@ 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)
+ ;
+ ; 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 +7475,30 @@ 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)
+ ;
+ ; 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 +7557,30 @@ 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)
+ ;
+ ; 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 +7639,30 @@ 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)
+ ;
+ ; 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 +7883,32 @@ 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)
+ ;
+ ; 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 +8031,32 @@ 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)
+ ;
+ ; 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 +8119,32 @@ 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)
+ ;
+ ; 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 +8207,32 @@ 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)
+ ;
+ ; 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 +8295,32 @@ 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)
+ ;
+ ; 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 +8383,32 @@ 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)
+ ;
+ ; 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 +8538,96 @@ 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)
+ ;
+ ; 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 +8690,32 @@ 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)
+ ;
+ ; 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 +8937,32 @@ 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)
+ ;
+ ; 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 +9085,32 @@ 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)
+ ;
+ ; 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 +9173,32 @@ 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)
+ ;
+ ; 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 +9261,32 @@ 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)
+ ;
+ ; 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 +9349,32 @@ 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)
+ ;
+ ; 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 +9685,28 @@ 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)
+ ;
+ ; 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 +9861,28 @@ 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)
+ ;
+ ; 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 +9937,28 @@ 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)
+ ;
+ ; 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 +10013,28 @@ 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)
+ ;
+ ; 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 +10089,28 @@ 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)
+ ;
+ ; 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 +10169,30 @@ 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)
+ ;
+ ; 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 +10507,30 @@ 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)
+ ;
+ ; 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 +10689,30 @@ 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)
+ ;
+ ; 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 +10771,30 @@ 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)
+ ;
+ ; 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 +10853,30 @@ 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)
+ ;
+ ; 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 +10935,30 @@ 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)
+ ;
+ ; 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 +11017,30 @@ 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)
+ ;
+ ; 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 +11095,28 @@ 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)
+ ;
+ ; 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 +11484,46 @@ 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)
+ ;
+ ; 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 +11719,46 @@ 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)
+ ;
+ ; 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 +11852,46 @@ 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)
+ ;
+ ; 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 +11985,46 @@ 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)
+ ;
+ ; 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 +12118,46 @@ 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)
+ ;
+ ; 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/load-d16.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
index 5d5b752b27a3c..d80ef5f4f4265 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
@@ -1,5 +1,7 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12 %s
+; FIXME-TRUE16
+; XFAIL: *
define amdgpu_ps void @load_P0_B16_D16(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
; GFX12-LABEL: load_P0_B16_D16:
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/flat-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
index a5576343af219..bc1b8faf5ecf2 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
@@ -2331,15 +2331,15 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-FAKE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v1
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
-; GFX1250-GISEL: ; %bb.0:
-; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3]
-; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
-; GFX1250-GISEL-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
+; GFX1250-GISEL-FAKE16: ; %bb.0:
+; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3]
+; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2349,6 +2349,16 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.h
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
+; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
+; GFX1250-GISEL-TRUE16: ; %bb.0:
+; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3]
+; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
@@ -2374,15 +2384,15 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-FAKE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v1
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
-; GFX1250-GISEL: ; %bb.0:
-; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
-; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
-; GFX1250-GISEL-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
+; GFX1250-GISEL-FAKE16: ; %bb.0:
+; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
+; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2392,6 +2402,16 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.h
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
+; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
+; GFX1250-GISEL-TRUE16: ; %bb.0:
+; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
+; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
@@ -2802,16 +2822,16 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-FAKE16-NEXT: v_perm_b32 v0, v0, v1, 0x5040100
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
-; GFX1250-GISEL: ; %bb.0:
-; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3]
-; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
-; GFX1250-GISEL-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
+; GFX1250-GISEL-FAKE16: ; %bb.0:
+; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3]
+; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1250-GISEL-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2823,6 +2843,17 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
+; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
+; GFX1250-GISEL-TRUE16: ; %bb.0:
+; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3]
+; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1250-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
@@ -2848,16 +2879,16 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-FAKE16-NEXT: v_perm_b32 v0, v0, v1, 0x5040100
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
-; GFX1250-GISEL: ; %bb.0:
-; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
-; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
-; GFX1250-GISEL-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
+; GFX1250-GISEL-FAKE16: ; %bb.0:
+; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
+; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1250-GISEL-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2869,6 +2900,17 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
+; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
+; GFX1250-GISEL-TRUE16: ; %bb.0:
+; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
+; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
+; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1250-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
+;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
diff --git a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
index 9e4c6e6935596..416ecca3515e5 100644
--- a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
@@ -5,7 +5,8 @@
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG,GFX12-SDAG-TRUE16 %s
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG,GFX12-SDAG-FAKE16 %s
-; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL,GFX12-GISEL-TRUE16 %s
+; FIXME:
+; xUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL,GFX12-GISEL-TRUE16 %s
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL,GFX12-GISEL-FAKE16 %s
; Test using saddr addressing mode of global_*load_* flat instructions.
@@ -4004,7 +4005,7 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16lo_zero_hi(ptr addrspace(1
; GFX12-GISEL-TRUE16: ; %bb.0:
; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3]
; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_cvt_u32_u16_e32 v0, v0.l
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16lo_zero_hi:
@@ -4050,7 +4051,7 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16lo_zero_hi_immneg128(ptr a
; GFX12-GISEL-TRUE16: ; %bb.0:
; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3] offset:-128
; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_cvt_u32_u16_e32 v0, v0.l
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16lo_zero_hi_immneg128:
@@ -4203,12 +4204,28 @@ 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
@@ -4233,12 +4250,28 @@ 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
@@ -4379,9 +4412,9 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16hi_zero_hi(ptr addrspace(1
;
; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16hi_zero_hi:
; GFX12-GISEL-TRUE16: ; %bb.0:
-; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3]
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_hi_b16 v0, v0, s[2:3]
; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16hi_zero_hi:
@@ -4441,9 +4474,9 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16hi_zero_hi_immneg128(ptr a
;
; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16hi_zero_hi_immneg128:
; GFX12-GISEL-TRUE16: ; %bb.0:
-; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3] offset:-128
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_hi_b16 v0, v0, s[2:3] offset:-128
; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16hi_zero_hi_immneg128:
@@ -4596,12 +4629,29 @@ 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
@@ -4626,12 +4676,29 @@ 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
>From 0a1b81495532aecbbffb9b2e735038bf81c90239 Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Mon, 27 Apr 2026 16:59:30 -0400
Subject: [PATCH 2/6] PR feedback, fix a few broken tests by adding to regbank
combiner
---
.../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 3 +-
.../Target/AMDGPU/AMDGPURegBankCombiner.cpp | 24 ++++
.../AMDGPU/AMDGPURegBankLegalizeRules.cpp | 7 +-
.../CodeGen/AMDGPU/GlobalISel/load-d16.ll | 2 -
llvm/test/CodeGen/AMDGPU/global-saddr-load.ll | 129 +++++-------------
5 files changed, 63 insertions(+), 102 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index b3afa80675a07..195d36e502239 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -1574,13 +1574,14 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
{S64, ConstantPtr, S64, GlobalAlign32},
{V2S32, ConstantPtr, V2S32, GlobalAlign32}});
- if (ST.useRealTrue16Insts())
+ 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 {
diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp
index 990d879c2bf09..fc9cbc760149b 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp
@@ -418,6 +418,18 @@ bool AMDGPURegBankCombinerImpl::combineD16Load(MachineInstr &MI) const {
return false;
}
+ if (Load->getOpcode() == AMDGPU::G_ZEXT) {
+ SextLoad = getDefIgnoringCopies(Load->getOperand(1).getReg(), MRI);
+ if (!SextLoad || SextLoad->getOpcode() != AMDGPU::G_SEXTLOAD)
+ return false;
+
+ const MachineMemOperand *MMO = *SextLoad->memoperands_begin();
+ if (MMO->getSizeInBits().getValue() != 8)
+ return false;
+
+ return applyD16Load(AMDGPU::G_AMDGPU_LOAD_D16_LO_I8, MI, SextLoad, Dst);
+ }
+
if (mi_match(
Load, MRI,
m_GAnd(m_MInstr(SextLoad), m_Copy(m_SpecificICst(CleanHi16))))) {
@@ -450,6 +462,18 @@ bool AMDGPURegBankCombinerImpl::combineD16Load(MachineInstr &MI) const {
return false;
}
+ if (Load->getOpcode() == AMDGPU::G_ZEXT) {
+ SextLoad = getDefIgnoringCopies(Load->getOperand(1).getReg(), MRI);
+ if (!SextLoad || SextLoad->getOpcode() != AMDGPU::G_SEXTLOAD)
+ return false;
+
+ const MachineMemOperand *MMO = *SextLoad->memoperands_begin();
+ if (MMO->getSizeInBits().getValue() != 8)
+ return false;
+
+ return applyD16Load(AMDGPU::G_AMDGPU_LOAD_D16_HI_I8, MI, SextLoad, Dst);
+ }
+
if (mi_match(
Load, MRI,
m_GAnd(m_MInstr(SextLoad), m_Copy(m_SpecificICst(CleanHi16))))) {
diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
index b5a82eb25b82b..4910cf0c50678 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
@@ -1107,23 +1107,28 @@ RegBankLegalizeRules::RegBankLegalizeRules(const GCNSubtarget &_ST,
addRulesForGOpcs({G_ZEXTLOAD, G_SEXTLOAD}) // i8 and i16 zeroextending loads
.Any({{DivS32, P0}, {{Vgpr32}, {VgprP0}}})
+ .Any({{DivS16, P0}, {{Vgpr16}, {VgprP0}}}, usesTrue16)
.Any({{DivS32, P1}, {{Vgpr32}, {VgprP1}}})
+ .Any({{DivS16, P1}, {{Vgpr16}, {VgprP1}}}, usesTrue16)
.Any({{{UniS32, P1}, isAlign4 && isUL}, {{Sgpr32}, {SgprP1}, WidenMMOToS32}}, !hasSMRDSmall)
.Any({{{UniS32, P1}, isNaturalAligned && isUL}, {{Sgpr32}, {SgprP1}}}, hasSMRDSmall)
.Any({{{UniS32, P1}, !isAlign4 || !isUL}, {{UniInVgprS32}, {SgprP1}}}, !hasSMRDSmall)
.Any({{{UniS32, P1}, !isNaturalAligned || !isUL}, {{UniInVgprS32}, {SgprP1}}}, hasSMRDSmall)
.Any({{DivS32, P3}, {{Vgpr32}, {VgprP3}}})
+ .Any({{DivS16, P3}, {{Vgpr16}, {VgprP3}}}, usesTrue16)
.Any({{UniS32, P3}, {{UniInVgprS32}, {VgprP3}}})
.Any({{DivS32, P4}, {{Vgpr32}, {VgprP4}}})
+ .Any({{DivS16, P4}, {{Vgpr16}, {VgprP4}}}, usesTrue16)
.Any({{{UniS32, P4}, isAlign4 && isUL}, {{Sgpr32}, {SgprP4}, WidenMMOToS32}}, !hasSMRDSmall)
.Any({{{UniS32, P4}, isNaturalAligned && isUL}, {{Sgpr32}, {SgprP4}}}, hasSMRDSmall)
.Any({{{UniS32, P4}, !isAlign4 || !isUL}, {{UniInVgprS32}, {SgprP4}}}, !hasSMRDSmall)
.Any({{{UniS32, P4}, !isNaturalAligned || !isUL}, {{UniInVgprS32}, {SgprP4}}}, hasSMRDSmall)
- .Any({{DivS32, P5}, {{Vgpr32}, {VgprP5}}});
+ .Any({{DivS32, P5}, {{Vgpr32}, {VgprP5}}})
+ .Any({{DivS16, P5}, {{Vgpr16}, {VgprP5}}}, usesTrue16);
addRulesForGOpcs({G_STORE})
// addrspace(0)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
index d80ef5f4f4265..5d5b752b27a3c 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
@@ -1,7 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12 %s
-; FIXME-TRUE16
-; XFAIL: *
define amdgpu_ps void @load_P0_B16_D16(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
; GFX12-LABEL: load_P0_B16_D16:
diff --git a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
index 416ecca3515e5..9e4c6e6935596 100644
--- a/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-saddr-load.ll
@@ -5,8 +5,7 @@
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG,GFX12-SDAG-TRUE16 %s
; RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG,GFX12-SDAG-FAKE16 %s
-; FIXME:
-; xUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL,GFX12-GISEL-TRUE16 %s
+; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=+real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL,GFX12-GISEL-TRUE16 %s
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -mattr=-real-true16 -mattr=+wavefrontsize64 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL,GFX12-GISEL-FAKE16 %s
; Test using saddr addressing mode of global_*load_* flat instructions.
@@ -4005,7 +4004,7 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16lo_zero_hi(ptr addrspace(1
; GFX12-GISEL-TRUE16: ; %bb.0:
; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 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: v_cvt_u32_u16_e32 v0, v0.l
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16lo_zero_hi:
@@ -4051,7 +4050,7 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16lo_zero_hi_immneg128(ptr a
; GFX12-GISEL-TRUE16: ; %bb.0:
; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 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: v_cvt_u32_u16_e32 v0, v0.l
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16lo_zero_hi_immneg128:
@@ -4204,28 +4203,12 @@ 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-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
+; 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
%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
@@ -4250,28 +4233,12 @@ 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-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
+; 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
%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
@@ -4412,9 +4379,9 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16hi_zero_hi(ptr addrspace(1
;
; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16hi_zero_hi:
; GFX12-GISEL-TRUE16: ; %bb.0:
-; GFX12-GISEL-TRUE16-NEXT: global_load_d16_hi_b16 v0, v0, s[2:3]
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3]
; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
+; GFX12-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16hi_zero_hi:
@@ -4474,9 +4441,9 @@ define amdgpu_ps <2 x half> @global_load_saddr_i16_d16hi_zero_hi_immneg128(ptr a
;
; GFX12-GISEL-TRUE16-LABEL: global_load_saddr_i16_d16hi_zero_hi_immneg128:
; GFX12-GISEL-TRUE16: ; %bb.0:
-; GFX12-GISEL-TRUE16-NEXT: global_load_d16_hi_b16 v0, v0, s[2:3] offset:-128
+; GFX12-GISEL-TRUE16-NEXT: global_load_d16_b16 v0, v0, s[2:3] offset:-128
; GFX12-GISEL-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
+; GFX12-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-GISEL-TRUE16-NEXT: ; return to shader part epilog
;
; GFX12-GISEL-FAKE16-LABEL: global_load_saddr_i16_d16hi_zero_hi_immneg128:
@@ -4629,29 +4596,12 @@ 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-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
+; 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
%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
@@ -4676,29 +4626,12 @@ 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-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
+; 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
%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
>From 56cc09dea6c44827335f128decad409b7a4f88e3 Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Mon, 27 Apr 2026 17:04:17 -0400
Subject: [PATCH 3/6] Combine mattrs in test
---
llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
index fc0f84b0594dc..7ca84036076ba 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
@@ -5,8 +5,8 @@
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds < %s | FileCheck -check-prefix=GFX11 %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=+real-true16 < %s | FileCheck -check-prefix=GFX12 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX9 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX10 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode,+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX9 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode,+enable-flat-scratch < %s | FileCheck -check-prefixes=UNALIGNED_GFX10 %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode < %s | FileCheck -check-prefixes=UNALIGNED_GFX942 %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=-unaligned-access-mode < %s | FileCheck -check-prefixes=UNALIGNED_GFX11 %s
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 -global-isel -new-reg-bank-select -disable-promote-alloca-to-vector -disable-promote-alloca-to-lds -mattr=+real-true16,-unaligned-access-mode < %s | FileCheck -check-prefixes=UNALIGNED_GFX12 %s
>From 0f7d6d2bd43e04c40db825c6668ec9ba8bdfd9a1 Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Fri, 1 May 2026 11:01:10 -0400
Subject: [PATCH 4/6] Revert some changes that will be made in separate PR.
Remove common case code that unsure if it even speeds up compile time.
Refactor combiner cases, add GFX10 testcase to ensure exist combines actually
do something.
---
.../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 15 +-
.../Target/AMDGPU/AMDGPURegBankCombiner.cpp | 43 ++-
.../CodeGen/AMDGPU/GlobalISel/load-d16.ll | 253 ++++++++++++++++++
3 files changed, 273 insertions(+), 38 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 195d36e502239..c8a20b978e5d2 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -464,8 +464,8 @@ static bool isLoadStoreSizeLegal(const GCNSubtarget &ST,
MemSize = std::max(MemSize, Align);
#endif
- // Only allow extloads to up to 32 bits.
- if (MemSize != RegSize && RegSize > 32)
+ // Only 1-byte and 2-byte to 32-bit extloads are valid.
+ if (MemSize != RegSize && RegSize != 32)
return false;
if (MemSize > maxSizeForAddrSpace(ST, AS, IsLoad,
@@ -544,7 +544,7 @@ static bool loadStoreBitcastWorkaround(const LLT Ty) {
static bool isLoadStoreLegal(const GCNSubtarget &ST, const LegalityQuery &Query) {
const LLT Ty = Query.Types[0];
- return isRegisterClassType(ST, Ty) && isLoadStoreSizeLegal(ST, Query) &&
+ return isRegisterType(ST, Ty) && isLoadStoreSizeLegal(ST, Query) &&
!hasBufferRsrcWorkaround(Ty) && !loadStoreBitcastWorkaround(Ty);
}
@@ -1574,15 +1574,6 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
{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);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp
index fc9cbc760149b..540730162e0cb 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegBankCombiner.cpp
@@ -418,21 +418,10 @@ bool AMDGPURegBankCombinerImpl::combineD16Load(MachineInstr &MI) const {
return false;
}
- if (Load->getOpcode() == AMDGPU::G_ZEXT) {
- SextLoad = getDefIgnoringCopies(Load->getOperand(1).getReg(), MRI);
- if (!SextLoad || SextLoad->getOpcode() != AMDGPU::G_SEXTLOAD)
- return false;
-
- const MachineMemOperand *MMO = *SextLoad->memoperands_begin();
- if (MMO->getSizeInBits().getValue() != 8)
- return false;
-
- return applyD16Load(AMDGPU::G_AMDGPU_LOAD_D16_LO_I8, MI, SextLoad, Dst);
- }
-
if (mi_match(
Load, MRI,
- m_GAnd(m_MInstr(SextLoad), m_Copy(m_SpecificICst(CleanHi16))))) {
+ m_GAnd(m_MInstr(SextLoad), m_Copy(m_SpecificICst(CleanHi16)))) ||
+ mi_match(Load, MRI, m_GZExt(m_MInstr(SextLoad)))) {
if (SextLoad->getOpcode() != AMDGPU::G_SEXTLOAD)
return false;
@@ -440,6 +429,12 @@ bool AMDGPURegBankCombinerImpl::combineD16Load(MachineInstr &MI) const {
if (MMO->getSizeInBits().getValue() != 8)
return false;
+ if (Load->getOpcode() == TargetOpcode::G_ZEXT) {
+ LLT SextDstTy = MRI.getType(SextLoad->getOperand(0).getReg());
+ if (SextDstTy.getSizeInBits() != 16)
+ return false;
+ }
+
return applyD16Load(AMDGPU::G_AMDGPU_LOAD_D16_LO_I8, MI, SextLoad, Dst);
}
@@ -462,27 +457,23 @@ bool AMDGPURegBankCombinerImpl::combineD16Load(MachineInstr &MI) const {
return false;
}
- if (Load->getOpcode() == AMDGPU::G_ZEXT) {
- SextLoad = getDefIgnoringCopies(Load->getOperand(1).getReg(), MRI);
- if (!SextLoad || SextLoad->getOpcode() != AMDGPU::G_SEXTLOAD)
- return false;
-
- const MachineMemOperand *MMO = *SextLoad->memoperands_begin();
- if (MMO->getSizeInBits().getValue() != 8)
- return false;
-
- return applyD16Load(AMDGPU::G_AMDGPU_LOAD_D16_HI_I8, MI, SextLoad, Dst);
- }
-
if (mi_match(
Load, MRI,
- m_GAnd(m_MInstr(SextLoad), m_Copy(m_SpecificICst(CleanHi16))))) {
+ m_GAnd(m_MInstr(SextLoad), m_Copy(m_SpecificICst(CleanHi16)))) ||
+ mi_match(Load, MRI, m_GZExt(m_MInstr(SextLoad)))) {
if (SextLoad->getOpcode() != AMDGPU::G_SEXTLOAD)
return false;
+
const MachineMemOperand *MMO = *SextLoad->memoperands_begin();
if (MMO->getSizeInBits().getValue() != 8)
return false;
+ if (Load->getOpcode() == TargetOpcode::G_ZEXT) {
+ LLT SextDstTy = MRI.getType(SextLoad->getOperand(0).getReg());
+ if (SextDstTy.getSizeInBits() != 16)
+ return false;
+ }
+
return applyD16Load(AMDGPU::G_AMDGPU_LOAD_D16_HI_I8, MI, SextLoad, Dst);
}
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
index 5d5b752b27a3c..977a9cda6128e 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/load-d16.ll
@@ -1,7 +1,15 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1030 < %s | FileCheck --check-prefixes=GFX10 %s
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck --check-prefixes=GFX12 %s
define amdgpu_ps void @load_P0_B16_D16(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
+; GFX10-LABEL: load_P0_B16_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: flat_load_short_d16 v0, v[1:2]
+; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT: flat_store_dword v[3:4], v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P0_B16_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: flat_load_d16_b16 v0, v[1:2]
@@ -15,6 +23,13 @@ define amdgpu_ps void @load_P0_B16_D16(<2 x i16> %vec, ptr addrspace(0) %ptra, p
}
define amdgpu_ps void @load_P0_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
+; GFX10-LABEL: load_P0_B16_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: flat_load_short_d16_hi v0, v[1:2]
+; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT: flat_store_dword v[3:4], v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P0_B16_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: flat_load_d16_hi_b16 v0, v[1:2]
@@ -28,6 +43,13 @@ define amdgpu_ps void @load_P0_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(0) %ptra
}
define amdgpu_ps void @sextload_P0_i8_D16(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
+; GFX10-LABEL: sextload_P0_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: flat_load_sbyte_d16 v0, v[1:2]
+; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT: flat_store_dword v[3:4], v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P0_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: flat_load_d16_i8 v0, v[1:2]
@@ -42,6 +64,13 @@ define amdgpu_ps void @sextload_P0_i8_D16(<2 x i16> %vec, ptr addrspace(0) %ptra
}
define amdgpu_ps void @sextload_P0_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
+; GFX10-LABEL: sextload_P0_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: flat_load_sbyte_d16_hi v0, v[1:2]
+; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT: flat_store_dword v[3:4], v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P0_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: flat_load_d16_hi_i8 v0, v[1:2]
@@ -56,6 +85,13 @@ define amdgpu_ps void @sextload_P0_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(0) %p
}
define amdgpu_ps void @zextload_P0_i8_D16(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
+; GFX10-LABEL: zextload_P0_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: flat_load_ubyte_d16 v0, v[1:2]
+; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT: flat_store_dword v[3:4], v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P0_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: flat_load_d16_u8 v0, v[1:2]
@@ -70,6 +106,13 @@ define amdgpu_ps void @zextload_P0_i8_D16(<2 x i16> %vec, ptr addrspace(0) %ptra
}
define amdgpu_ps void @zextload_P0_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(0) %ptra, ptr addrspace(0) %out) {
+; GFX10-LABEL: zextload_P0_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: flat_load_ubyte_d16_hi v0, v[1:2]
+; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX10-NEXT: flat_store_dword v[3:4], v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P0_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: flat_load_d16_hi_u8 v0, v[1:2]
@@ -84,6 +127,13 @@ define amdgpu_ps void @zextload_P0_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(0) %p
}
define amdgpu_ps void @load_P1_B16_D16(<2 x i16> %vec, ptr addrspace(1) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: load_P1_B16_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_short_d16 v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P1_B16_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_b16 v0, v[1:2], off
@@ -97,6 +147,13 @@ define amdgpu_ps void @load_P1_B16_D16(<2 x i16> %vec, ptr addrspace(1) %ptra, p
}
define amdgpu_ps void @load_P1_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(1) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: load_P1_B16_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_short_d16_hi v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P1_B16_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_hi_b16 v0, v[1:2], off
@@ -110,6 +167,13 @@ define amdgpu_ps void @load_P1_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(1) %ptra
}
define amdgpu_ps void @sextload_P1_i8_D16(<2 x i16> %vec, ptr addrspace(1) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: sextload_P1_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_sbyte_d16 v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P1_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_i8 v0, v[1:2], off
@@ -124,6 +188,13 @@ define amdgpu_ps void @sextload_P1_i8_D16(<2 x i16> %vec, ptr addrspace(1) %ptra
}
define amdgpu_ps void @sextload_P1_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(1) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: sextload_P1_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_sbyte_d16_hi v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P1_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_hi_i8 v0, v[1:2], off
@@ -138,6 +209,13 @@ define amdgpu_ps void @sextload_P1_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(1) %p
}
define amdgpu_ps void @zextload_P1_i8_D16(<2 x i16> %vec, ptr addrspace(1) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: zextload_P1_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_ubyte_d16 v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P1_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_u8 v0, v[1:2], off
@@ -152,6 +230,13 @@ define amdgpu_ps void @zextload_P1_i8_D16(<2 x i16> %vec, ptr addrspace(1) %ptra
}
define amdgpu_ps void @zextload_P1_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(1) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: zextload_P1_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_ubyte_d16_hi v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P1_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_hi_u8 v0, v[1:2], off
@@ -166,6 +251,13 @@ define amdgpu_ps void @zextload_P1_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(1) %p
}
define amdgpu_ps void @load_P3_B16_D16(<2 x i16> %vec, ptr addrspace(3) %ptra, ptr addrspace(3) %out) {
+; GFX10-LABEL: load_P3_B16_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: ds_read_u16_d16 v0, v1
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: ds_write_b32 v2, v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P3_B16_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: ds_load_u16_d16 v0, v1
@@ -179,6 +271,13 @@ define amdgpu_ps void @load_P3_B16_D16(<2 x i16> %vec, ptr addrspace(3) %ptra, p
}
define amdgpu_ps void @load_P3_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(3) %ptra, ptr addrspace(3) %out) {
+; GFX10-LABEL: load_P3_B16_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: ds_read_u16_d16_hi v0, v1
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: ds_write_b32 v2, v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P3_B16_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: ds_load_u16_d16_hi v0, v1
@@ -192,6 +291,13 @@ define amdgpu_ps void @load_P3_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(3) %ptra
}
define amdgpu_ps void @sextload_P3_i8_D16(<2 x i16> %vec, ptr addrspace(3) %ptra, ptr addrspace(3) %out) {
+; GFX10-LABEL: sextload_P3_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: ds_read_i8_d16 v0, v1
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: ds_write_b32 v2, v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P3_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: ds_load_i8_d16 v0, v1
@@ -206,6 +312,13 @@ define amdgpu_ps void @sextload_P3_i8_D16(<2 x i16> %vec, ptr addrspace(3) %ptra
}
define amdgpu_ps void @sextload_P3_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(3) %ptra, ptr addrspace(3) %out) {
+; GFX10-LABEL: sextload_P3_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: ds_read_i8_d16_hi v0, v1
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: ds_write_b32 v2, v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P3_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: ds_load_i8_d16_hi v0, v1
@@ -220,6 +333,13 @@ define amdgpu_ps void @sextload_P3_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(3) %p
}
define amdgpu_ps void @zextload_P3_i8_D16(<2 x i16> %vec, ptr addrspace(3) %ptra, ptr addrspace(3) %out) {
+; GFX10-LABEL: zextload_P3_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: ds_read_u8_d16 v0, v1
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: ds_write_b32 v2, v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P3_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: ds_load_u8_d16 v0, v1
@@ -234,6 +354,13 @@ define amdgpu_ps void @zextload_P3_i8_D16(<2 x i16> %vec, ptr addrspace(3) %ptra
}
define amdgpu_ps void @zextload_P3_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(3) %ptra, ptr addrspace(3) %out) {
+; GFX10-LABEL: zextload_P3_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: ds_read_u8_d16_hi v0, v1
+; GFX10-NEXT: s_waitcnt lgkmcnt(0)
+; GFX10-NEXT: ds_write_b32 v2, v0
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P3_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: ds_load_u8_d16_hi v0, v1
@@ -248,6 +375,13 @@ define amdgpu_ps void @zextload_P3_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(3) %p
}
define amdgpu_ps void @load_P4_B16_D16(<2 x i16> %vec, ptr addrspace(4) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: load_P4_B16_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_short_d16 v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P4_B16_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_b16 v0, v[1:2], off
@@ -261,6 +395,13 @@ define amdgpu_ps void @load_P4_B16_D16(<2 x i16> %vec, ptr addrspace(4) %ptra, p
}
define amdgpu_ps void @load_P4_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(4) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: load_P4_B16_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_short_d16_hi v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P4_B16_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_hi_b16 v0, v[1:2], off
@@ -274,6 +415,13 @@ define amdgpu_ps void @load_P4_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(4) %ptra
}
define amdgpu_ps void @sextload_P4_i8_D16(<2 x i16> %vec, ptr addrspace(4) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: sextload_P4_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_sbyte_d16 v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P4_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_i8 v0, v[1:2], off
@@ -288,6 +436,13 @@ define amdgpu_ps void @sextload_P4_i8_D16(<2 x i16> %vec, ptr addrspace(4) %ptra
}
define amdgpu_ps void @sextload_P4_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(4) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: sextload_P4_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_sbyte_d16_hi v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P4_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_hi_i8 v0, v[1:2], off
@@ -302,6 +457,13 @@ define amdgpu_ps void @sextload_P4_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(4) %p
}
define amdgpu_ps void @zextload_P4_i8_D16(<2 x i16> %vec, ptr addrspace(4) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: zextload_P4_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_ubyte_d16 v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P4_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_u8 v0, v[1:2], off
@@ -316,6 +478,13 @@ define amdgpu_ps void @zextload_P4_i8_D16(<2 x i16> %vec, ptr addrspace(4) %ptra
}
define amdgpu_ps void @zextload_P4_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(4) %ptra, ptr addrspace(1) %out) {
+; GFX10-LABEL: zextload_P4_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: global_load_ubyte_d16_hi v0, v[1:2], off
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: global_store_dword v[3:4], v0, off
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P4_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: global_load_d16_hi_u8 v0, v[1:2], off
@@ -330,6 +499,20 @@ define amdgpu_ps void @zextload_P4_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(4) %p
}
define amdgpu_ps void @load_P5_B16_D16(<2 x i16> %vec, ptr addrspace(5) %ptra, ptr addrspace(5) %out) {
+; GFX10-LABEL: load_P5_B16_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_mov_b32 s4, s0
+; GFX10-NEXT: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
+; GFX10-NEXT: s_mov_b32 s1, SCRATCH_RSRC_DWORD1
+; GFX10-NEXT: s_mov_b32 s2, -1
+; GFX10-NEXT: s_mov_b32 s3, 0x31c16000
+; GFX10-NEXT: s_add_u32 s0, s0, s4
+; GFX10-NEXT: s_addc_u32 s1, s1, 0
+; GFX10-NEXT: buffer_load_short_d16 v0, v1, s[0:3], 0 offen
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P5_B16_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: scratch_load_d16_b16 v0, v1, off
@@ -343,6 +526,20 @@ define amdgpu_ps void @load_P5_B16_D16(<2 x i16> %vec, ptr addrspace(5) %ptra, p
}
define amdgpu_ps void @load_P5_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(5) %ptra, ptr addrspace(5) %out) {
+; GFX10-LABEL: load_P5_B16_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_mov_b32 s4, s0
+; GFX10-NEXT: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
+; GFX10-NEXT: s_mov_b32 s1, SCRATCH_RSRC_DWORD1
+; GFX10-NEXT: s_mov_b32 s2, -1
+; GFX10-NEXT: s_mov_b32 s3, 0x31c16000
+; GFX10-NEXT: s_add_u32 s0, s0, s4
+; GFX10-NEXT: s_addc_u32 s1, s1, 0
+; GFX10-NEXT: buffer_load_short_d16_hi v0, v1, s[0:3], 0 offen
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: load_P5_B16_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: scratch_load_d16_hi_b16 v0, v1, off
@@ -356,6 +553,20 @@ define amdgpu_ps void @load_P5_B16_D16_Hi(<2 x i16> %vec, ptr addrspace(5) %ptra
}
define amdgpu_ps void @sextload_P5_i8_D16(<2 x i16> %vec, ptr addrspace(5) %ptra, ptr addrspace(5) %out) {
+; GFX10-LABEL: sextload_P5_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_mov_b32 s4, s0
+; GFX10-NEXT: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
+; GFX10-NEXT: s_mov_b32 s1, SCRATCH_RSRC_DWORD1
+; GFX10-NEXT: s_mov_b32 s2, -1
+; GFX10-NEXT: s_mov_b32 s3, 0x31c16000
+; GFX10-NEXT: s_add_u32 s0, s0, s4
+; GFX10-NEXT: s_addc_u32 s1, s1, 0
+; GFX10-NEXT: buffer_load_sbyte_d16 v0, v1, s[0:3], 0 offen
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P5_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: scratch_load_d16_i8 v0, v1, off
@@ -370,6 +581,20 @@ define amdgpu_ps void @sextload_P5_i8_D16(<2 x i16> %vec, ptr addrspace(5) %ptra
}
define amdgpu_ps void @sextload_P5_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(5) %ptra, ptr addrspace(5) %out) {
+; GFX10-LABEL: sextload_P5_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_mov_b32 s4, s0
+; GFX10-NEXT: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
+; GFX10-NEXT: s_mov_b32 s1, SCRATCH_RSRC_DWORD1
+; GFX10-NEXT: s_mov_b32 s2, -1
+; GFX10-NEXT: s_mov_b32 s3, 0x31c16000
+; GFX10-NEXT: s_add_u32 s0, s0, s4
+; GFX10-NEXT: s_addc_u32 s1, s1, 0
+; GFX10-NEXT: buffer_load_sbyte_d16_hi v0, v1, s[0:3], 0 offen
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: sextload_P5_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: scratch_load_d16_hi_i8 v0, v1, off
@@ -384,6 +609,20 @@ define amdgpu_ps void @sextload_P5_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(5) %p
}
define amdgpu_ps void @zextload_P5_i8_D16(<2 x i16> %vec, ptr addrspace(5) %ptra, ptr addrspace(5) %out) {
+; GFX10-LABEL: zextload_P5_i8_D16:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_mov_b32 s4, s0
+; GFX10-NEXT: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
+; GFX10-NEXT: s_mov_b32 s1, SCRATCH_RSRC_DWORD1
+; GFX10-NEXT: s_mov_b32 s2, -1
+; GFX10-NEXT: s_mov_b32 s3, 0x31c16000
+; GFX10-NEXT: s_add_u32 s0, s0, s4
+; GFX10-NEXT: s_addc_u32 s1, s1, 0
+; GFX10-NEXT: buffer_load_ubyte_d16 v0, v1, s[0:3], 0 offen
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P5_i8_D16:
; GFX12: ; %bb.0:
; GFX12-NEXT: scratch_load_d16_u8 v0, v1, off
@@ -398,6 +637,20 @@ define amdgpu_ps void @zextload_P5_i8_D16(<2 x i16> %vec, ptr addrspace(5) %ptra
}
define amdgpu_ps void @zextload_P5_i8_D16_Hi(<2 x i16> %vec, ptr addrspace(5) %ptra, ptr addrspace(5) %out) {
+; GFX10-LABEL: zextload_P5_i8_D16_Hi:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: s_mov_b32 s4, s0
+; GFX10-NEXT: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
+; GFX10-NEXT: s_mov_b32 s1, SCRATCH_RSRC_DWORD1
+; GFX10-NEXT: s_mov_b32 s2, -1
+; GFX10-NEXT: s_mov_b32 s3, 0x31c16000
+; GFX10-NEXT: s_add_u32 s0, s0, s4
+; GFX10-NEXT: s_addc_u32 s1, s1, 0
+; GFX10-NEXT: buffer_load_ubyte_d16_hi v0, v1, s[0:3], 0 offen
+; GFX10-NEXT: s_waitcnt vmcnt(0)
+; GFX10-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
+; GFX10-NEXT: s_endpgm
+;
; GFX12-LABEL: zextload_P5_i8_D16_Hi:
; GFX12: ; %bb.0:
; GFX12-NEXT: scratch_load_d16_hi_u8 v0, v1, off
>From 5963d3d9928c8311c5683fa4f6a6a29a85236ec3 Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Fri, 1 May 2026 11:26:54 -0400
Subject: [PATCH 5/6] Update testing after splitting out changes
---
.../CodeGen/AMDGPU/GlobalISel/flat-scratch.ll | 162 +++++++++---------
.../AMDGPU/GlobalISel/store-local.128.ll | 64 ++++---
.../AMDGPU/GlobalISel/store-local.96.ll | 48 +++---
llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll | 118 ++++---------
4 files changed, 166 insertions(+), 226 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
index 7ca84036076ba..27385fb316545 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
@@ -2689,23 +2689,22 @@ 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_mov_b16_e32 v1.l, 15
-; UNALIGNED_GFX11-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX11-NEXT: v_dual_mov_b32 v1, 15 :: v_dual_mov_b32 v2, 0
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:1 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:2 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:3 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, 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: scratch_store_d16_hi_b8 v0, v1, off offset:5 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:5 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:6 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, 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)
@@ -2732,24 +2731,23 @@ define void @store_load_i64_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 15
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 15 :: v_dual_mov_b32 v2, 0
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
@@ -3030,33 +3028,32 @@ 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_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: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
+; UNALIGNED_GFX11-NEXT: v_mov_b32_e32 v3, 2
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:1 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:2 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:3 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; 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: v_mov_b32_e32 v1, 3
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v3, off offset:4 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:5 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:6 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, 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_d16_hi_b8 v0, v1, off offset:9 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:9 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:10 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, 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)
@@ -3091,34 +3088,33 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 1
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 2
+; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 2
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 3
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v1, 3
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off offset:8 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:9 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:10 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:11 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
@@ -3462,42 +3458,41 @@ 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_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: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
+; UNALIGNED_GFX11-NEXT: v_mov_b32_e32 v3, 2
; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v1, off dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:1 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:2 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, 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_b8 v0, v3, off offset:4 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; 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: 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: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:6 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, 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_d16_hi_b8 v0, v1, off offset:9 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:9 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:10 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:11 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:12 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v3, off offset:12 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:13 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:13 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:14 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, off offset:14 dlc
; UNALIGNED_GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; UNALIGNED_GFX11-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:15 dlc
+; UNALIGNED_GFX11-NEXT: scratch_store_b8 v0, v2, 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)
@@ -3540,43 +3535,42 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 1
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 2
+; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 2
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 3
-; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 4
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v1, 3
+; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 4
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off offset:8 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:9 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:10 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:11 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:12 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:12 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:13 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:13 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:14 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:14 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:15 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:15 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll
index ec4f3efa2d150..10e83b70a57d4 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.128.ll
@@ -239,53 +239,47 @@ 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_mov_b16_e32 v0.l, s0
+; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s4
; 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_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_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: 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: s_lshr_b32 s0, s7, 8
+; GFX11-NEXT: v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s6
; GFX11-NEXT: s_and_b32 s9, 0xffff, s2
-; GFX11-NEXT: v_mov_b16_e32 v4.l, s0
+; GFX11-NEXT: v_dual_mov_b32 v8, s1 :: v_dual_mov_b32 v9, 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 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: 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: s_and_b32 s0, 0xffff, s3
; GFX11-NEXT: s_lshr_b32 s8, s2, 16
-; GFX11-NEXT: v_mov_b16_e32 v1.l, s2
-; GFX11-NEXT: v_mov_b16_e32 v0.l, s1
+; GFX11-NEXT: v_dual_mov_b32 v0, s1 :: v_dual_mov_b32 v5, s3
; 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: v_mov_b16_e32 v2.h, s0
+; 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: s_lshr_b32 s0, s3, 24
-; 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: 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: 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 158b5d7450609..db5d0fdb5c036 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.96.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/store-local.96.ll
@@ -206,43 +206,37 @@ 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 s4, s0, 16
+; GFX11-NEXT: s_lshr_b32 s7, s2, 16
; GFX11-NEXT: s_lshr_b32 s5, s5, 8
-; GFX11-NEXT: v_mov_b16_e32 v0.l, s0
+; GFX11-NEXT: s_lshr_b32 s4, s0, 16
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: v_mov_b32_e32 v6, s3
+; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, 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_mov_b16_e32 v0.h, s1
+; GFX11-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v3, s2
; GFX11-NEXT: s_lshr_b32 s1, s1, 24
; GFX11-NEXT: s_and_b32 s8, 0xffff, s2
-; 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: 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: 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_mov_b16_e32 v3.l, s1
+; GFX11-NEXT: v_dual_mov_b32 v6, s3 :: v_dual_mov_b32 v7, s1
; GFX11-NEXT: s_lshr_b32 s1, s8, 8
-; 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: 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: s_endpgm
store <3 x i32> %x, ptr addrspace(3) %out, align 1
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
index bc1b8faf5ecf2..a5576343af219 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
@@ -2331,15 +2331,15 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-FAKE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v1
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
-; GFX1250-GISEL-FAKE16: ; %bb.0:
-; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3]
-; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
-; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
+; GFX1250-GISEL: ; %bb.0:
+; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3]
+; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX1250-GISEL-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2349,16 +2349,6 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.h
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
-; GFX1250-GISEL-TRUE16: ; %bb.0:
-; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3]
-; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
-; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
-; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
-;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
@@ -2384,15 +2374,15 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-FAKE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v1
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
-; GFX1250-GISEL-FAKE16: ; %bb.0:
-; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
-; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
-; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
+; GFX1250-GISEL: ; %bb.0:
+; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
+; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
+; GFX1250-GISEL-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2402,16 +2392,6 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.h
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
-; GFX1250-GISEL-TRUE16: ; %bb.0:
-; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
-; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
-; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff0000, v1, v0
-; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
-;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16lo_sexti8_reg_hi_immneg128:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
@@ -2822,16 +2802,16 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-FAKE16-NEXT: v_perm_b32 v0, v0, v1, 0x5040100
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
-; GFX1250-GISEL-FAKE16: ; %bb.0:
-; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3]
-; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
-; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
+; GFX1250-GISEL: ; %bb.0:
+; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3]
+; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1250-GISEL-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX1250-GISEL-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2843,17 +2823,6 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi(ptr inreg %
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
-; GFX1250-GISEL-TRUE16: ; %bb.0:
-; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3]
-; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
-; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
-; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
-;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
@@ -2879,16 +2848,16 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-FAKE16-NEXT: v_perm_b32 v0, v0, v1, 0x5040100
; GFX1250-SDAG-FAKE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-FAKE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
-; GFX1250-GISEL-FAKE16: ; %bb.0:
-; GFX1250-GISEL-FAKE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-FAKE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
-; GFX1250-GISEL-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1250-GISEL-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-GISEL-FAKE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
-; GFX1250-GISEL-FAKE16-NEXT: ; return to shader part epilog
+; GFX1250-GISEL-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
+; GFX1250-GISEL: ; %bb.0:
+; GFX1250-GISEL-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
+; GFX1250-GISEL-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
+; GFX1250-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
+; GFX1250-GISEL-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1250-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1250-GISEL-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-GISEL-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
+; GFX1250-GISEL-NEXT: ; return to shader part epilog
;
; GFX1250-SDAG-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
; GFX1250-SDAG-TRUE16: ; %bb.0:
@@ -2900,17 +2869,6 @@ define amdgpu_ps <2 x half> @flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128(p
; GFX1250-SDAG-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1250-SDAG-TRUE16-NEXT: ; return to shader part epilog
;
-; GFX1250-GISEL-TRUE16-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
-; GFX1250-GISEL-TRUE16: ; %bb.0:
-; GFX1250-GISEL-TRUE16-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
-; GFX1250-GISEL-TRUE16-NEXT: flat_load_i8 v0, v0, s[2:3] offset:-128
-; GFX1250-GISEL-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0
-; GFX1250-GISEL-TRUE16-NEXT: v_mov_b16_e32 v0.h, 0
-; GFX1250-GISEL-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-GISEL-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-GISEL-TRUE16-NEXT: v_and_or_b32 v0, 0xffff, v1, v0
-; GFX1250-GISEL-TRUE16-NEXT: ; return to shader part epilog
-;
; GFX1250-NOECC-LABEL: flat_load_saddr_i16_d16hi_sexti8_reg_hi_immneg128:
; GFX1250-NOECC: ; %bb.0:
; GFX1250-NOECC-NEXT: s_setreg_imm32_b32 hwreg(HW_REG_WAVE_MODE, 25, 1), 1 ; msbs: dst=0 src0=0 src1=0 src2=0
>From 24beae3e5f670702000303a50d29f42423dd4886 Mon Sep 17 00:00:00 2001
From: Domenic Nutile <domenic.nutile at gmail.com>
Date: Mon, 4 May 2026 12:09:53 -0400
Subject: [PATCH 6/6] Readd change and testing update mistakenly removed
---
.../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 9 +
.../CodeGen/AMDGPU/GlobalISel/flat-scratch.ll | 162 +++++++++---------
.../AMDGPU/GlobalISel/store-local.128.ll | 64 +++----
.../AMDGPU/GlobalISel/store-local.96.ll | 48 +++---
4 files changed, 155 insertions(+), 128 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index c8a20b978e5d2..6fee3ea94409d 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -1574,6 +1574,15 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
{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);
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
index 27385fb316545..7ca84036076ba 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/flat-scratch.ll
@@ -2689,22 +2689,23 @@ 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)
@@ -2731,23 +2732,24 @@ define void @store_load_i64_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 15 :: v_dual_mov_b32 v2, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 15
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
@@ -3028,32 +3030,33 @@ 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)
@@ -3088,33 +3091,34 @@ define void @store_load_v3i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 2
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 1
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 2
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v1, 3
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 3
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off offset:8 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:9 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:10 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:11 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
@@ -3458,41 +3462,42 @@ 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)
@@ -3535,42 +3540,43 @@ define void @store_load_v4i32_unaligned(ptr addrspace(5) nocapture %arg) {
; UNALIGNED_GFX12-NEXT: s_wait_samplecnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_bvhcnt 0x0
; UNALIGNED_GFX12-NEXT: s_wait_kmcnt 0x0
-; UNALIGNED_GFX12-NEXT: v_dual_mov_b32 v1, 1 :: v_dual_mov_b32 v2, 0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 2
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 1
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.h, 0
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 2
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:1 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:1 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:2 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:2 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:3 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:3 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:4 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:4 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v1, 3
-; UNALIGNED_GFX12-NEXT: v_mov_b32_e32 v3, 4
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:5 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v1.l, 3
+; UNALIGNED_GFX12-NEXT: v_mov_b16_e32 v2.l, 4
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:5 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:6 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:6 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:7 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:7 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v1, off offset:8 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:9 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:9 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:10 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:10 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:11 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:11 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v3, off offset:12 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:12 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:13 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:13 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:14 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:14 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
-; UNALIGNED_GFX12-NEXT: scratch_store_b8 v0, v2, off offset:15 scope:SCOPE_SYS
+; UNALIGNED_GFX12-NEXT: scratch_store_d16_hi_b8 v0, v1, off offset:15 scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_storecnt 0x0
; UNALIGNED_GFX12-NEXT: scratch_load_u8 v1, v0, off scope:SCOPE_SYS
; UNALIGNED_GFX12-NEXT: s_wait_loadcnt 0x0
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
More information about the llvm-commits
mailing list