[llvm] Revert "Update amdgpu_gfx functions to use s0-s3 for inreg SGPR arguments on targets using scratch instructions for stack #78226" (PR #86273)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Mar 22 05:22:38 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-globalisel
Author: Thomas Symalla (tsymalla)
<details>
<summary>Changes</summary>
Reverts llvm/llvm-project#<!-- -->81394
This reverts commit 3ac243bc0d7922d083af2cf025247b5698556062.
It is not handling RSrc registers s0-s3 correctly. This leads to a
broken test, where it expects s0-s3 as function argument and uses it as
RSrc register as well.
We need to re-visit the patch, but apparently we only want to have s0-s3 as
argument registers if we don't need them as RSrc registers.
---
Patch is 343.28 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/86273.diff
12 Files Affected:
- (modified) llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp (+1-4)
- (modified) llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td (-1)
- (modified) llvm/lib/Target/AMDGPU/SIISelLowering.cpp (+3-4)
- (modified) llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-non-fixed.ll (+5-5)
- (modified) llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll (+5-5)
- (modified) llvm/test/CodeGen/AMDGPU/bf16.ll (+1-1)
- (modified) llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll (+12-12)
- (modified) llvm/test/CodeGen/AMDGPU/function-args-inreg.ll (-133)
- (modified) llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll (+2503-1874)
- (modified) llvm/test/CodeGen/AMDGPU/indirect-call.ll (+30-30)
- (modified) llvm/test/CodeGen/AMDGPU/schedule-addrspaces.ll (+1-1)
- (modified) llvm/test/CodeGen/AMDGPU/scratch-pointer-sink.ll (+2-2)
``````````diff
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
index 8969b3b5b6ce08..7e1f041fa10933 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
@@ -715,13 +715,10 @@ bool AMDGPUCallLowering::lowerFormalArguments(
if (!IsEntryFunc && !IsGraphics) {
// For the fixed ABI, pass workitem IDs in the last argument register.
TLI.allocateSpecialInputVGPRsFixed(CCInfo, MF, *TRI, *Info);
- }
- if (!IsEntryFunc) {
if (!Subtarget.enableFlatScratch())
CCInfo.AllocateReg(Info->getScratchRSrcReg());
- if (!IsGraphics)
- TLI.allocateSpecialInputSGPRs(CCInfo, MF, *TRI, *Info);
+ TLI.allocateSpecialInputSGPRs(CCInfo, MF, *TRI, *Info);
}
IncomingValueAssigner Assigner(AssignFn);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td b/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
index 9bd0fd7eca76fb..4be64629ddac82 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
@@ -23,7 +23,6 @@ def CC_SI_Gfx : CallingConv<[
// 33 is reserved for the frame pointer
// 34 is reserved for the base pointer
CCIfInReg<CCIfType<[f32, i32, f16, i16, v2i16, v2f16, bf16, v2bf16] , CCAssignToReg<[
- SGPR0, SGPR1, SGPR2, SGPR3,
SGPR4, SGPR5, SGPR6, SGPR7,
SGPR8, SGPR9, SGPR10, SGPR11, SGPR12, SGPR13, SGPR14, SGPR15,
SGPR16, SGPR17, SGPR18, SGPR19, SGPR20, SGPR21, SGPR22, SGPR23,
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 7f0cff72c18661..5ccf21f76015de 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -2857,13 +2857,12 @@ SDValue SITargetLowering::LowerFormalArguments(
} else if (!IsGraphics) {
// For the fixed ABI, pass workitem IDs in the last argument register.
allocateSpecialInputVGPRsFixed(CCInfo, MF, *TRI, *Info);
- }
- if (!IsEntryFunc) {
+ // FIXME: Sink this into allocateSpecialInputSGPRs
if (!Subtarget->enableFlatScratch())
CCInfo.AllocateReg(Info->getScratchRSrcReg());
- if (!IsGraphics)
- allocateSpecialInputSGPRs(CCInfo, MF, *TRI, *Info);
+
+ allocateSpecialInputSGPRs(CCInfo, MF, *TRI, *Info);
}
if (!IsKernel) {
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-non-fixed.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-non-fixed.ll
index fad833c0a6ad5e..5effd24a752088 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-non-fixed.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-non-fixed.ll
@@ -50,10 +50,10 @@ define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm_inreg(i32 inreg
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
- ; CHECK-NEXT: $sgpr0 = COPY [[C]](s32)
+ ; CHECK-NEXT: $sgpr4 = COPY [[C]](s32)
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
- ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+ ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
; CHECK-NEXT: SI_RETURN
call amdgpu_gfx void @external_gfx_void_func_i32_inreg(i32 inreg 42)
@@ -99,11 +99,11 @@ define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32_inreg() #
; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
- ; CHECK-NEXT: $sgpr0 = COPY [[ANYEXT1]](s32)
- ; CHECK-NEXT: $sgpr1 = COPY [[LOAD2]](s32)
+ ; CHECK-NEXT: $sgpr4 = COPY [[ANYEXT1]](s32)
+ ; CHECK-NEXT: $sgpr5 = COPY [[LOAD2]](s32)
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
- ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr0, implicit $sgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+ ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
; CHECK-NEXT: SI_RETURN
%ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll
index 75670604baa1ac..392b0ae6823e44 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll
@@ -942,10 +942,10 @@ define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm_inreg(i32 inreg
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc
; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
- ; CHECK-NEXT: $sgpr0 = COPY [[C]](s32)
+ ; CHECK-NEXT: $sgpr4 = COPY [[C]](s32)
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
- ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+ ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
; CHECK-NEXT: SI_RETURN
call amdgpu_gfx void @external_gfx_void_func_i32_inreg(i32 inreg 42)
@@ -3984,11 +3984,11 @@ define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32_inreg() #
; CHECK-NEXT: [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
- ; CHECK-NEXT: $sgpr0 = COPY [[ANYEXT1]](s32)
- ; CHECK-NEXT: $sgpr1 = COPY [[LOAD2]](s32)
+ ; CHECK-NEXT: $sgpr4 = COPY [[ANYEXT1]](s32)
+ ; CHECK-NEXT: $sgpr5 = COPY [[LOAD2]](s32)
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]](<4 x s32>)
- ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr0, implicit $sgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+ ; CHECK-NEXT: $sgpr30_sgpr31 = noconvergent G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_si_gfx, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc
; CHECK-NEXT: SI_RETURN
%ptr0 = load ptr addrspace(1), ptr addrspace(4) undef
diff --git a/llvm/test/CodeGen/AMDGPU/bf16.ll b/llvm/test/CodeGen/AMDGPU/bf16.ll
index e369f7e3b9a5ad..98658834e89784 100644
--- a/llvm/test/CodeGen/AMDGPU/bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/bf16.ll
@@ -3337,7 +3337,7 @@ define amdgpu_gfx void @test_inreg_arg_store(bfloat inreg %in, ptr addrspace(1)
; GFX11-LABEL: test_inreg_arg_store:
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_mov_b32_e32 v2, s0
+; GFX11-NEXT: v_mov_b32_e32 v2, s4
; GFX11-NEXT: global_store_b16 v[0:1], v2, off
; GFX11-NEXT: s_setpc_b64 s[30:31]
store bfloat %in, ptr addrspace(1) %out
diff --git a/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll b/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll
index e1e3220cc27554..10d71a315fbf98 100644
--- a/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll
+++ b/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll
@@ -472,7 +472,7 @@ define amdgpu_gfx void @test34(i32 inreg %arg1, i32 inreg %arg2) {
; GCN-LABEL: test34:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_min_i32 s0, s0, s1
+; GCN-NEXT: s_min_i32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
; GCN-NEXT: s_cmpk_lt_i32 s0, 0x3e9
; GCN-NEXT: v_mov_b32_e32 v1, 0
@@ -492,7 +492,7 @@ define amdgpu_gfx void @test35(i32 inreg %arg1, i32 inreg %arg2) {
; GCN-LABEL: test35:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_max_i32 s0, s0, s1
+; GCN-NEXT: s_max_i32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
; GCN-NEXT: s_cmpk_gt_i32 s0, 0x3e8
; GCN-NEXT: v_mov_b32_e32 v1, 0
@@ -512,9 +512,9 @@ define amdgpu_gfx void @test36(i32 inreg %arg1, i32 inreg %arg2, i32 inreg %arg3
; GCN-LABEL: test36:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_min_u32 s0, s0, s1
+; GCN-NEXT: s_min_u32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
-; GCN-NEXT: s_cmp_lt_u32 s0, s2
+; GCN-NEXT: s_cmp_lt_u32 s0, s6
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: s_cselect_b32 s0, -1, 0
; GCN-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0
@@ -532,9 +532,9 @@ define amdgpu_gfx void @test37(i32 inreg %arg1, i32 inreg %arg2, i32 inreg %arg3
; GCN-LABEL: test37:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_max_i32 s0, s0, s1
+; GCN-NEXT: s_max_i32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
-; GCN-NEXT: s_cmp_ge_i32 s0, s2
+; GCN-NEXT: s_cmp_ge_i32 s0, s6
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: s_cselect_b32 s0, -1, 0
; GCN-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0
@@ -552,7 +552,7 @@ define amdgpu_gfx void @test38(i32 inreg %arg1, i32 inreg %arg2) {
; GCN-LABEL: test38:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_max_u32 s0, s0, s1
+; GCN-NEXT: s_max_u32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
; GCN-NEXT: s_cmpk_lt_u32 s0, 0x3e9
; GCN-NEXT: v_mov_b32_e32 v1, 0
@@ -572,7 +572,7 @@ define amdgpu_gfx void @test39(i32 inreg %arg1, i32 inreg %arg2) {
; GCN-LABEL: test39:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_min_i32 s0, s0, s1
+; GCN-NEXT: s_min_i32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
; GCN-NEXT: s_cmpk_gt_i32 s0, 0x3e7
; GCN-NEXT: v_mov_b32_e32 v1, 0
@@ -592,9 +592,9 @@ define amdgpu_gfx void @test40(i32 inreg %arg1, i32 inreg %arg2, i32 inreg %arg3
; GCN-LABEL: test40:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_max_i32 s0, s0, s1
+; GCN-NEXT: s_max_i32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
-; GCN-NEXT: s_cmp_le_i32 s0, s2
+; GCN-NEXT: s_cmp_le_i32 s0, s6
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: s_cselect_b32 s0, -1, 0
; GCN-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0
@@ -612,9 +612,9 @@ define amdgpu_gfx void @test41(i32 inreg %arg1, i32 inreg %arg2, i32 inreg %arg3
; GCN-LABEL: test41:
; GCN: ; %bb.0:
; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT: s_min_u32 s0, s0, s1
+; GCN-NEXT: s_min_u32 s0, s4, s5
; GCN-NEXT: v_mov_b32_e32 v0, 0
-; GCN-NEXT: s_cmp_ge_u32 s0, s2
+; GCN-NEXT: s_cmp_ge_u32 s0, s6
; GCN-NEXT: v_mov_b32_e32 v1, 0
; GCN-NEXT: s_cselect_b32 s0, -1, 0
; GCN-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0
diff --git a/llvm/test/CodeGen/AMDGPU/function-args-inreg.ll b/llvm/test/CodeGen/AMDGPU/function-args-inreg.ll
index 27845b6b5b2fee..44a9127b4bd09c 100644
--- a/llvm/test/CodeGen/AMDGPU/function-args-inreg.ll
+++ b/llvm/test/CodeGen/AMDGPU/function-args-inreg.ll
@@ -2176,93 +2176,6 @@ define void @void_func_a5i32_inreg([5 x i32] inreg %arg0, ptr addrspace(1) %ptr)
declare void @extern()
define void @void_func_a13i32_inreg([13 x i32] inreg %arg0, ptr addrspace(1) %ptr) {
-; GFX9-LABEL: void_func_a13i32_inreg:
-; GFX9: ; %bb.0:
-; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: s_mov_b32 s27, s33
-; GFX9-NEXT: s_mov_b32 s33, s32
-; GFX9-NEXT: s_or_saveexec_b64 s[28:29], -1
-; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
-; GFX9-NEXT: s_mov_b64 exec, s[28:29]
-; GFX9-NEXT: v_mov_b32_e32 v2, s26
-; GFX9-NEXT: global_store_dword v[0:1], v2, off offset:48
-; GFX9-NEXT: v_mov_b32_e32 v5, s25
-; GFX9-NEXT: v_mov_b32_e32 v4, s24
-; GFX9-NEXT: v_mov_b32_e32 v3, s23
-; GFX9-NEXT: v_mov_b32_e32 v2, s22
-; GFX9-NEXT: s_addk_i32 s32, 0x400
-; GFX9-NEXT: global_store_dwordx4 v[0:1], v[2:5], off offset:32
-; GFX9-NEXT: v_writelane_b32 v40, s27, 2
-; GFX9-NEXT: v_mov_b32_e32 v5, s21
-; GFX9-NEXT: v_mov_b32_e32 v4, s20
-; GFX9-NEXT: v_mov_b32_e32 v3, s19
-; GFX9-NEXT: v_mov_b32_e32 v2, s18
-; GFX9-NEXT: global_store_dwordx4 v[0:1], v[2:5], off offset:16
-; GFX9-NEXT: v_writelane_b32 v40, s30, 0
-; GFX9-NEXT: v_mov_b32_e32 v5, s17
-; GFX9-NEXT: v_mov_b32_e32 v4, s16
-; GFX9-NEXT: s_getpc_b64 s[16:17]
-; GFX9-NEXT: s_add_u32 s16, s16, extern at gotpcrel32@lo+4
-; GFX9-NEXT: s_addc_u32 s17, s17, extern at gotpcrel32@hi+12
-; GFX9-NEXT: s_load_dwordx2 s[16:17], s[16:17], 0x0
-; GFX9-NEXT: v_mov_b32_e32 v3, s7
-; GFX9-NEXT: v_mov_b32_e32 v2, s6
-; GFX9-NEXT: v_writelane_b32 v40, s31, 1
-; GFX9-NEXT: global_store_dwordx4 v[0:1], v[2:5], off
-; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX9-NEXT: v_readlane_b32 s31, v40, 1
-; GFX9-NEXT: v_readlane_b32 s30, v40, 0
-; GFX9-NEXT: v_readlane_b32 s4, v40, 2
-; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1
-; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
-; GFX9-NEXT: s_mov_b64 exec, s[6:7]
-; GFX9-NEXT: s_addk_i32 s32, 0xfc00
-; GFX9-NEXT: s_mov_b32 s33, s4
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: s_setpc_b64 s[30:31]
-;
-; GFX11-LABEL: void_func_a13i32_inreg:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: s_mov_b32 s23, s33
-; GFX11-NEXT: s_mov_b32 s33, s32
-; GFX11-NEXT: s_or_saveexec_b32 s24, -1
-; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill
-; GFX11-NEXT: s_mov_b32 exec_lo, s24
-; GFX11-NEXT: s_add_i32 s32, s32, 16
-; GFX11-NEXT: v_dual_mov_b32 v4, s20 :: v_dual_mov_b32 v3, s19
-; GFX11-NEXT: v_dual_mov_b32 v2, s18 :: v_dual_mov_b32 v9, s17
-; GFX11-NEXT: s_getpc_b64 s[18:19]
-; GFX11-NEXT: s_add_u32 s18, s18, extern at gotpcrel32@lo+4
-; GFX11-NEXT: s_addc_u32 s19, s19, extern at gotpcrel32@hi+12
-; GFX11-NEXT: v_dual_mov_b32 v8, s16 :: v_dual_mov_b32 v7, s7
-; GFX11-NEXT: s_load_b64 s[16:17], s[18:19], 0x0
-; GFX11-NEXT: v_writelane_b32 v40, s23, 2
-; GFX11-NEXT: v_dual_mov_b32 v14, s22 :: v_dual_mov_b32 v5, s21
-; GFX11-NEXT: v_dual_mov_b32 v6, s6 :: v_dual_mov_b32 v13, s3
-; GFX11-NEXT: v_writelane_b32 v40, s30, 0
-; GFX11-NEXT: v_dual_mov_b32 v12, s2 :: v_dual_mov_b32 v11, s1
-; GFX11-NEXT: v_mov_b32_e32 v10, s0
-; GFX11-NEXT: s_clause 0x3
-; GFX11-NEXT: global_store_b32 v[0:1], v14, off offset:48
-; GFX11-NEXT: global_store_b128 v[0:1], v[2:5], off offset:32
-; GFX11-NEXT: global_store_b128 v[0:1], v[6:9], off offset:16
-; GFX11-NEXT: global_store_b128 v[0:1], v[10:13], off
-; GFX11-NEXT: v_writelane_b32 v40, s31, 1
-; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT: v_readlane_b32 s31, v40, 1
-; GFX11-NEXT: v_readlane_b32 s30, v40, 0
-; GFX11-NEXT: v_readlane_b32 s0, v40, 2
-; GFX11-NEXT: s_or_saveexec_b32 s1, -1
-; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
-; GFX11-NEXT: s_mov_b32 exec_lo, s1
-; GFX11-NEXT: s_add_i32 s32, s32, -16
-; GFX11-NEXT: s_mov_b32 s33, s0
-; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: s_setpc_b64 s[30:31]
store [13 x i32] %arg0, ptr addrspace(1) %ptr
call void @extern()
ret void
@@ -2290,52 +2203,6 @@ define void @void_func_a13i32_inreg([13 x i32] inreg %arg0, ptr addrspace(1) %p
; FIXME: Should still fail
define void @void_func_a16i32_inreg__noimplicit([16 x i32] inreg %arg0, ptr addrspace(1) %ptr) {
-; GFX9-LABEL: void_func_a16i32_inreg__noimplicit:
-; GFX9: ; %bb.0:
-; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v5, s19
-; GFX9-NEXT: v_mov_b32_e32 v4, s18
-; GFX9-NEXT: v_mov_b32_e32 v3, s17
-; GFX9-NEXT: v_mov_b32_e32 v2, s16
-; GFX9-NEXT: global_store_dwordx4 v[0:1], v[2:5], off offset:48
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: v_mov_b32_e32 v5, s15
-; GFX9-NEXT: v_mov_b32_e32 v4, s14
-; GFX9-NEXT: v_mov_b32_e32 v3, s13
-; GFX9-NEXT: v_mov_b32_e32 v2, s12
-; GFX9-NEXT: global_store_dwordx4 v[0:1], v[2:5], off offset:32
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: v_mov_b32_e32 v5, s11
-; GFX9-NEXT: v_mov_b32_e32 v4, s10
-; GFX9-NEXT: v_mov_b32_e32 v3, s9
-; GFX9-NEXT: v_mov_b32_e32 v2, s8
-; GFX9-NEXT: global_store_dwordx4 v[0:1], v[2:5], off offset:16
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: v_mov_b32_e32 v5, s7
-; GFX9-NEXT: v_mov_b32_e32 v4, s6
-; GFX9-NEXT: v_mov_b32_e32 v3, s5
-; GFX9-NEXT: v_mov_b32_e32 v2, s4
-; GFX9-NEXT: global_store_dwordx4 v[0:1], v[2:5], off
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: s_setpc_b64 s[30:31]
-;
-; GFX11-LABEL: void_func_a16i32_inreg__noimplicit:
-; GFX11: ; %bb.0:
-; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-NEXT: v_dual_mov_b32 v5, s15 :: v_dual_mov_b32 v4, s14
-; GFX11-NEXT: v_dual_mov_b32 v3, s13 :: v_dual_mov_b32 v2, s12
-; GFX11-NEXT: v_dual_mov_b32 v9, s11 :: v_dual_mov_b32 v8, s10
-; GFX11-NEXT: v_dual_mov_b32 v7, s9 :: v_dual_mov_b32 v6, s8
-; GFX11-NEXT: v_dual_mov_b32 v13, s7 :: v_dual_mov_b32 v12, s6
-; GFX11-NEXT: v_dual_mov_b32 v11, s5 :: v_dual_mov_b32 v10, s4
-; GFX11-NEXT: v_dual_mov_b32 v17, s3 :: v_dual_mov_b32 v16, s2
-; GFX11-NEXT: v_dual_mov_b32 v15, s1 :: v_dual_mov_b32 v14, s0
-; GFX11-NEXT: s_clause 0x3
-; GFX11-NEXT: global_store_b128 v[0:1], v[2:5], off offset:48
-; GFX11-NEXT: global_store_b128 v[0:1], v[6:9], off offset:32
-; GFX11-NEXT: global_store_b128 v[0:1], v[10:13], off offset:16
-; GFX11-NEXT: global_store_b128 v[0:1], v[14:17], off
-; GFX11-NEXT: s_setpc_b64 s[30:31]
store [16 x i32] %arg0, ptr addrspace(1) %ptr
ret void
}
diff --git a/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll b/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
index 3e1db5fb4e1dc8..a118fa388f86d9 100644
--- a/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
+++ b/llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
@@ -9567,17 +9567,19 @@ define amdgpu_gfx void @test_call_external_void_func_i8_imm_inreg(i32) #0 {
; GFX9-NEXT: s_or_saveexec_b64 s[36:37], -1
; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GFX9-NEXT: s_mov_b64 exec, s[36:37]
-; GFX9-NEXT: v_writelane_b32 v40, s34, 2
-; GFX9-NEXT: v_writelane_b32 v40, s30, 0
+; GFX9-NEXT: v_writelane_b32 v40, s34, 3
+; GFX9-NEXT: v_writelane_b32 v40, s4, 0
+; GFX9-NEXT: v_writelane_b32 v40, s30, 1
; GFX9-NEXT: s_mov_b32 s35, external_void_func_i8_inreg at abs32@hi
; GFX9-NEXT: s_mov_b32 s34, external_void_func_i8_inreg at abs32@lo
-; GFX9-NEXT: s_movk_i32 s0, 0x7b
+; GFX9-NEXT: s_movk_i32 s4, 0x7b
; GFX9-NEXT: s_addk_i32 s32, 0x400
-; GFX9-NEXT: v_writelane_b32 v40, s31, 1
+; GFX9-NEXT: v_writelane_b32 v40, s31, 2
; GFX9-NEXT: s_swappc_b64 s[30:31], s[34:35]
-; GFX9-NEXT: v_readlane_b32 s31, v40, 1
-; GFX9-NEXT: v_readlane_b32 s30, v40, 0
-; GFX9-NEXT: v_readlane_b32 s34, v40, 2
+; GFX9-NEXT: ...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/86273
More information about the llvm-commits
mailing list