https://github.com/arsenm updated
https://github.com/llvm/llvm-project/pull/131101
>From 054d80a8ef3d5d456b2ff692d9bf7215c31933a8 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <matthew.arsena...@amd.com>
Date: Thu, 13 Mar 2025 14:55:45 +0700
Subject: [PATCH] AMDGPU: Replace test uses of ptr addrspace(5) undef with
poison
---
.../AMDGPU/GlobalISel/divergent-control-flow.ll | 2 +-
llvm/test/CodeGen/AMDGPU/collapse-endcf.ll | 4 ++--
llvm/test/CodeGen/AMDGPU/dag-divergence.ll | 2 +-
llvm/test/CodeGen/AMDGPU/debug-value.ll | 2 +-
llvm/test/CodeGen/AMDGPU/frame-index-elimination.ll | 2 +-
llvm/test/CodeGen/AMDGPU/lds-dma-waitcnt.mir | 2 +-
llvm/test/CodeGen/AMDGPU/mdt-preserving-crash.ll | 2 +-
...emory-legalizer-multiple-mem-operands-atomics.mir | 6 +++---
llvm/test/CodeGen/AMDGPU/move-to-valu-worklist.ll | 2 +-
llvm/test/CodeGen/AMDGPU/operand-folding.ll | 2 +-
.../test/CodeGen/AMDGPU/private-access-no-objects.ll | 4 ++--
.../CodeGen/AMDGPU/promote-alloca-to-lds-select.ll | 6 +++---
llvm/test/CodeGen/AMDGPU/sad.ll | 12 ++++++------
llvm/test/CodeGen/AMDGPU/scalar_to_vector_v2x16.ll | 8 ++++----
.../stack-pointer-offset-relative-frameindex.ll | 2 +-
15 files changed, 29 insertions(+), 29 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/divergent-control-flow.ll
b/llvm/test/CodeGen/AMDGPU/GlobalISel/divergent-control-flow.ll
index 3ad5845467cd0..989ee80a1f002 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/divergent-control-flow.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/divergent-control-flow.ll
@@ -183,7 +183,7 @@ bb8:
br i1 %tmp10, label %bb11, label %bb12
bb11:
- store float 4.0, ptr addrspace(5) undef, align 4
+ store float 4.0, ptr addrspace(5) poison, align 4
br label %bb12
bb12:
diff --git a/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
b/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
index c8a4f2d2e6b7b..a60a16c9be47e 100644
--- a/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
+++ b/llvm/test/CodeGen/AMDGPU/collapse-endcf.ll
@@ -1360,7 +1360,7 @@ bb2: ; preds
= %bb1
br i1 %tmp3, label %bb4, label %bb10
bb4: ; preds = %bb2
- %tmp6 = load float, ptr addrspace(5) undef
+ %tmp6 = load float, ptr addrspace(5) poison
%tmp7 = fcmp olt float %tmp6, 0.0
br i1 %tmp7, label %bb8, label %Flow
@@ -1380,7 +1380,7 @@ Flow1: ; preds
= %bb10
br label %bb1
bb12: ; preds = %bb10
- store volatile <4 x float> %tmp11, ptr addrspace(5) undef, align 16
+ store volatile <4 x float> %tmp11, ptr addrspace(5) poison, align 16
ret void
}
diff --git a/llvm/test/CodeGen/AMDGPU/dag-divergence.ll
b/llvm/test/CodeGen/AMDGPU/dag-divergence.ll
index 8b27a9bde8ea6..9f83393d88061 100644
--- a/llvm/test/CodeGen/AMDGPU/dag-divergence.ll
+++ b/llvm/test/CodeGen/AMDGPU/dag-divergence.ll
@@ -6,7 +6,7 @@
; GCN: flat_load_dword
; GCN-NOT: s_load_dword s
define amdgpu_kernel void @private_load_maybe_divergent(ptr addrspace(4) %k,
ptr %flat) {
- %load = load volatile i32, ptr addrspace(5) undef, align 4
+ %load = load volatile i32, ptr addrspace(5) poison, align 4
%gep = getelementptr inbounds i32, ptr addrspace(4) %k, i32 %load
%maybe.not.uniform.load = load i32, ptr addrspace(4) %gep, align 4
store i32 %maybe.not.uniform.load, ptr addrspace(1) poison
diff --git a/llvm/test/CodeGen/AMDGPU/debug-value.ll
b/llvm/test/CodeGen/AMDGPU/debug-value.ll
index 167c0ce7ceefa..f13bd665cc7f0 100644
--- a/llvm/test/CodeGen/AMDGPU/debug-value.ll
+++ b/llvm/test/CodeGen/AMDGPU/debug-value.ll
@@ -35,7 +35,7 @@ bb25: ; preds =
%bb
bb28: ; preds = %bb25, %bb21
%tmp29 = phi <4 x float> [ %tmp27, %bb25 ], [ %tmp24, %bb21 ]
- store <4 x float> %tmp29, ptr addrspace(5) undef, align 16
+ store <4 x float> %tmp29, ptr addrspace(5) poison, align 16
%tmp30 = getelementptr inbounds %struct.wombat, ptr addrspace(1) %arg, i64
%tmp2, i32 2, i64 2
%tmp31 = load i32, ptr addrspace(1) %tmp30, align 4
%tmp32 = sext i32 %tmp31 to i64
diff --git a/llvm/test/CodeGen/AMDGPU/frame-index-elimination.ll
b/llvm/test/CodeGen/AMDGPU/frame-index-elimination.ll
index ea33925117aea..25fe57c16c661 100644
--- a/llvm/test/CodeGen/AMDGPU/frame-index-elimination.ll
+++ b/llvm/test/CodeGen/AMDGPU/frame-index-elimination.ll
@@ -255,7 +255,7 @@ define void @undefined_stack_store_reg(float %arg, i32
%arg1) #0 {
bb:
%tmp = alloca <4 x float>, align 16, addrspace(5)
%tmp2 = insertelement <4 x float> poison, float %arg, i32 0
- store <4 x float> %tmp2, ptr addrspace(5) undef
+ store <4 x float> %tmp2, ptr addrspace(5) poison
%tmp3 = icmp eq i32 %arg1, 0
br i1 %tmp3, label %bb4, label %bb5
diff --git a/llvm/test/CodeGen/AMDGPU/lds-dma-waitcnt.mir
b/llvm/test/CodeGen/AMDGPU/lds-dma-waitcnt.mir
index f10b5378a7462..cc150b40d0c7e 100644
--- a/llvm/test/CodeGen/AMDGPU/lds-dma-waitcnt.mir
+++ b/llvm/test/CodeGen/AMDGPU/lds-dma-waitcnt.mir
@@ -77,7 +77,7 @@ name: scratch_load_lds_dword_ds_read
body: |
bb.0:
$m0 = S_MOV_B32 0
- SCRATCH_LOAD_LDS_DWORD $vgpr0, 4, 0, implicit $exec, implicit $m0 :: (load
(s32) from `ptr addrspace(5) undef` + 4), (store (s32) into `ptr addrspace(3)
undef` + 4)
+ SCRATCH_LOAD_LDS_DWORD $vgpr0, 4, 0, implicit $exec, implicit $m0 :: (load
(s32) from `ptr addrspace(5) poison` + 4), (store (s32) into `ptr addrspace(3)
undef` + 4)
$vgpr0 = DS_READ_B32_gfx9 $vgpr1, 0, 0, implicit $m0, implicit $exec ::
(load (s32) from `ptr addrspace(3) undef`)
S_ENDPGM 0
diff --git a/llvm/test/CodeGen/AMDGPU/mdt-preserving-crash.ll
b/llvm/test/CodeGen/AMDGPU/mdt-preserving-crash.ll
index 3bd8700e33661..ef1cbd78d7cd7 100644
--- a/llvm/test/CodeGen/AMDGPU/mdt-preserving-crash.ll
+++ b/llvm/test/CodeGen/AMDGPU/mdt-preserving-crash.ll
@@ -104,7 +104,7 @@ if.end5.i362: ; preds =
%if.then3.i356, %if.
%conv612.i359 = sext i8 %5 to i32
%sub13.i360 = add nsw i32 %conv612.i359, -48
%cmp714.i361 = icmp ugt i32 %sub13.i360, 9
- store i8 0, ptr addrspace(5) undef, align 16
+ store i8 0, ptr addrspace(5) poison, align 16
%6 = load i8, ptr addrspace(1) getelementptr inbounds ([4096 x i8], ptr
addrspace(1) @_RSENC_gDcd_______________________________, i64 0, i64 1153),
align 1
%arrayidx232250.1 = getelementptr inbounds [128 x i8], ptr addrspace(5)
%pD10, i32 0, i32 1
store i8 %6, ptr addrspace(5) %arrayidx232250.1, align 1
diff --git
a/llvm/test/CodeGen/AMDGPU/memory-legalizer-multiple-mem-operands-atomics.mir
b/llvm/test/CodeGen/AMDGPU/memory-legalizer-multiple-mem-operands-atomics.mir
index 6705d1bdb4bb0..ceb2033965e8f 100644
---
a/llvm/test/CodeGen/AMDGPU/memory-legalizer-multiple-mem-operands-atomics.mir
+++
b/llvm/test/CodeGen/AMDGPU/memory-legalizer-multiple-mem-operands-atomics.mir
@@ -23,13 +23,13 @@ body: |
$sgpr10 = S_MOV_B32 4294967295, implicit-def $sgpr8_sgpr9_sgpr10_sgpr11
$sgpr11 = S_MOV_B32 15204352, implicit-def $sgpr8_sgpr9_sgpr10_sgpr11
$vgpr0 = V_MOV_B32_e32 1, implicit $exec
- BUFFER_STORE_DWORD_OFFSET killed $vgpr0, $sgpr8_sgpr9_sgpr10_sgpr11,
$sgpr3, 4, 0, 0, implicit $exec :: (store (s32) into `ptr addrspace(5) undef`)
+ BUFFER_STORE_DWORD_OFFSET killed $vgpr0, $sgpr8_sgpr9_sgpr10_sgpr11,
$sgpr3, 4, 0, 0, implicit $exec :: (store (s32) into `ptr addrspace(5) poison`)
S_WAITCNT 127
S_CMP_LG_U32 killed $sgpr2, 0, implicit-def $scc
S_WAITCNT 3855
$vgpr0 = V_MOV_B32_e32 2, implicit $exec
$vgpr1 = V_MOV_B32_e32 32772, implicit $exec
- BUFFER_STORE_DWORD_OFFEN killed $vgpr0, killed $vgpr1,
$sgpr8_sgpr9_sgpr10_sgpr11, $sgpr3, 0, 0, 0, implicit $exec :: (store (s32)
into `ptr addrspace(5) undef`)
+ BUFFER_STORE_DWORD_OFFEN killed $vgpr0, killed $vgpr1,
$sgpr8_sgpr9_sgpr10_sgpr11, $sgpr3, 0, 0, 0, implicit $exec :: (store (s32)
into `ptr addrspace(5) poison`)
S_CBRANCH_SCC0 %bb.1, implicit killed $scc
bb.2:
@@ -55,7 +55,7 @@ body: |
S_WAITCNT 127
$sgpr0 = S_LSHL_B32 killed $sgpr0, 2, implicit-def dead $scc
$vgpr0 = V_ADD_CO_U32_e32 killed $sgpr0, killed $vgpr0, implicit-def dead
$vcc, implicit $exec
- $vgpr0 = BUFFER_LOAD_DWORD_OFFEN killed $vgpr0, killed
$sgpr8_sgpr9_sgpr10_sgpr11, $sgpr3, 0, 0, 0, implicit $exec :: (load
syncscope("agent-one-as") unordered (s32) from `ptr addrspace(1) undef`), (load
syncscope("workgroup-one-as") seq_cst (s32) from `ptr addrspace(5) undef`)
+ $vgpr0 = BUFFER_LOAD_DWORD_OFFEN killed $vgpr0, killed
$sgpr8_sgpr9_sgpr10_sgpr11, $sgpr3, 0, 0, 0, implicit $exec :: (load
syncscope("agent-one-as") unordered (s32) from `ptr addrspace(1) undef`), (load
syncscope("workgroup-one-as") seq_cst (s32) from `ptr addrspace(5) poison`)
$vgpr1 = V_MOV_B32_e32 $sgpr4, implicit $exec, implicit-def $vgpr1_vgpr2,
implicit $sgpr4_sgpr5
$vgpr2 = V_MOV_B32_e32 killed $sgpr5, implicit $exec, implicit
$sgpr4_sgpr5, implicit $exec
S_WAITCNT 3952
diff --git a/llvm/test/CodeGen/AMDGPU/move-to-valu-worklist.ll
b/llvm/test/CodeGen/AMDGPU/move-to-valu-worklist.ll
index a4cb43e14c2e2..a487650fe8ef2 100644
--- a/llvm/test/CodeGen/AMDGPU/move-to-valu-worklist.ll
+++ b/llvm/test/CodeGen/AMDGPU/move-to-valu-worklist.ll
@@ -13,7 +13,7 @@
; GCN-NEXT: v_and_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
define amdgpu_kernel void @in_worklist_once() #0 {
bb:
- %tmp = load i64, ptr addrspace(5) undef
+ %tmp = load i64, ptr addrspace(5) poison
br label %bb1
bb1: ; preds = %bb1, %bb
diff --git a/llvm/test/CodeGen/AMDGPU/operand-folding.ll
b/llvm/test/CodeGen/AMDGPU/operand-folding.ll
index 8bb10d5c7d20f..93631ff12e277 100644
--- a/llvm/test/CodeGen/AMDGPU/operand-folding.ll
+++ b/llvm/test/CodeGen/AMDGPU/operand-folding.ll
@@ -130,7 +130,7 @@ define amdgpu_kernel void @no_fold_tied_subregister() #1 {
; CHECK: v_xor_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
define void @no_extra_fold_on_same_opnd() #1 {
entry:
- %s0 = load i32, ptr addrspace(5) undef, align 4
+ %s0 = load i32, ptr addrspace(5) poison, align 4
%s0.i64= zext i32 %s0 to i64
br label %for.body.i.i
diff --git a/llvm/test/CodeGen/AMDGPU/private-access-no-objects.ll
b/llvm/test/CodeGen/AMDGPU/private-access-no-objects.ll
index f1e5e68927be7..e687ad940862c 100644
--- a/llvm/test/CodeGen/AMDGPU/private-access-no-objects.ll
+++ b/llvm/test/CodeGen/AMDGPU/private-access-no-objects.ll
@@ -18,7 +18,7 @@
; OPTNONE-NOT: s_mov_b32
; OPTNONE: buffer_store_dword v{{[0-9]+}}, v{{[0-9]+}}, s[0:3], 0 offen{{$}}
define amdgpu_kernel void @store_to_undef() #0 {
- store volatile i32 0, ptr addrspace(5) undef
+ store volatile i32 0, ptr addrspace(5) poison
ret void
}
@@ -36,7 +36,7 @@ define amdgpu_kernel void @store_to_inttoptr() #0 {
; OPT-DAG: s_mov_b64 s{{\[[0-9]+}}:[[RSRC_HI:[0-9]+]]], s[2:3]
; OPT: buffer_load_dword v{{[0-9]+}}, v{{[0-9]+}}, s[[[RSRC_LO]]:[[RSRC_HI]]],
0 offen glc{{$}}
define amdgpu_kernel void @load_from_undef() #0 {
- %ld = load volatile i32, ptr addrspace(5) undef
+ %ld = load volatile i32, ptr addrspace(5) poison
ret void
}
diff --git a/llvm/test/CodeGen/AMDGPU/promote-alloca-to-lds-select.ll
b/llvm/test/CodeGen/AMDGPU/promote-alloca-to-lds-select.ll
index 4a15dad069c63..4c2dd7956a032 100644
--- a/llvm/test/CodeGen/AMDGPU/promote-alloca-to-lds-select.ll
+++ b/llvm/test/CodeGen/AMDGPU/promote-alloca-to-lds-select.ll
@@ -4,10 +4,10 @@ target datalayout =
"e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:3
; CHECK-LABEL: @lds_promoted_alloca_select_invalid_pointer_operand(
; CHECK: %alloca = alloca i32
-; CHECK: select i1 undef, ptr addrspace(5) undef, ptr addrspace(5) %alloca
+; CHECK: select i1 undef, ptr addrspace(5) poison, ptr addrspace(5) %alloca
define amdgpu_kernel void
@lds_promoted_alloca_select_invalid_pointer_operand() #0 {
%alloca = alloca i32, align 4, addrspace(5)
- %select = select i1 undef, ptr addrspace(5) undef, ptr addrspace(5) %alloca
+ %select = select i1 undef, ptr addrspace(5) poison, ptr addrspace(5) %alloca
store i32 0, ptr addrspace(5) %select, align 4
ret void
}
@@ -87,7 +87,7 @@ entry:
bb1:
%ptr2 = getelementptr inbounds [16 x i32], ptr addrspace(5) %alloca, i32 0,
i32 %c
- %select0 = select i1 undef, ptr addrspace(5) undef, ptr addrspace(5) %ptr2
+ %select0 = select i1 undef, ptr addrspace(5) poison, ptr addrspace(5) %ptr2
store i32 0, ptr addrspace(5) %ptr1
br label %bb2
diff --git a/llvm/test/CodeGen/AMDGPU/sad.ll b/llvm/test/CodeGen/AMDGPU/sad.ll
index 27ab9ace811eb..5474338514522 100644
--- a/llvm/test/CodeGen/AMDGPU/sad.ll
+++ b/llvm/test/CodeGen/AMDGPU/sad.ll
@@ -105,7 +105,7 @@ define amdgpu_kernel void @v_sad_u32_multi_use_sub_pat1(ptr
addrspace(1) %out, i
%t1 = select i1 %icmp1, i32 %a, i32 %b
%ret0 = sub i32 %t0, %t1
- store volatile i32 %ret0, ptr addrspace(5) undef
+ store volatile i32 %ret0, ptr addrspace(5) poison
%ret = add i32 %ret0, %c
store i32 %ret, ptr addrspace(1) %out
@@ -139,7 +139,7 @@ define amdgpu_kernel void @v_sad_u32_multi_use_add_pat1(ptr
addrspace(1) %out, i
%ret0 = sub i32 %t0, %t1
%ret = add i32 %ret0, %c
- store volatile i32 %ret, ptr addrspace(5) undef
+ store volatile i32 %ret, ptr addrspace(5) poison
store i32 %ret, ptr addrspace(1) %out
ret void
}
@@ -167,7 +167,7 @@ define amdgpu_kernel void @v_sad_u32_multi_use_max_pat1(ptr
addrspace(1) %out, i
; GCN-NEXT: s_endpgm
%icmp0 = icmp ugt i32 %a, %b
%t0 = select i1 %icmp0, i32 %a, i32 %b
- store volatile i32 %t0, ptr addrspace(5) undef
+ store volatile i32 %t0, ptr addrspace(5) poison
%icmp1 = icmp ule i32 %a, %b
%t1 = select i1 %icmp1, i32 %a, i32 %b
@@ -206,7 +206,7 @@ define amdgpu_kernel void @v_sad_u32_multi_use_min_pat1(ptr
addrspace(1) %out, i
%icmp1 = icmp ule i32 %a, %b
%t1 = select i1 %icmp1, i32 %a, i32 %b
- store volatile i32 %t1, ptr addrspace(5) undef
+ store volatile i32 %t1, ptr addrspace(5) poison
%ret0 = sub i32 %t0, %t1
%ret = add i32 %ret0, %c
@@ -238,7 +238,7 @@ define amdgpu_kernel void @v_sad_u32_multi_use_sub_pat2(ptr
addrspace(1) %out, i
; GCN-NEXT: s_endpgm
%icmp0 = icmp ugt i32 %a, %b
%sub0 = sub i32 %a, %b
- store volatile i32 %sub0, ptr addrspace(5) undef
+ store volatile i32 %sub0, ptr addrspace(5) poison
%sub1 = sub i32 %b, %a
%ret0 = select i1 %icmp0, i32 %sub0, i32 %sub1
@@ -274,7 +274,7 @@ define amdgpu_kernel void
@v_sad_u32_multi_use_select_pat2(ptr addrspace(1) %out
%sub0 = sub i32 %a, %b
%sub1 = sub i32 %b, %a
%ret0 = select i1 %icmp0, i32 %sub0, i32 %sub1
- store volatile i32 %ret0, ptr addrspace(5) undef
+ store volatile i32 %ret0, ptr addrspace(5) poison
%ret = add i32 %ret0, %c
diff --git a/llvm/test/CodeGen/AMDGPU/scalar_to_vector_v2x16.ll
b/llvm/test/CodeGen/AMDGPU/scalar_to_vector_v2x16.ll
index 47e4406762872..b37a66d3be1ee 100644
--- a/llvm/test/CodeGen/AMDGPU/scalar_to_vector_v2x16.ll
+++ b/llvm/test/CodeGen/AMDGPU/scalar_to_vector_v2x16.ll
@@ -5,9 +5,9 @@
; GCN: v_mov_b32_e32 [[V:v[0-9]+]], 42
; GCN: buffer_store_short [[V]],
define void @scalar_to_vector_i16() {
- %tmp = load <2 x i16>, ptr addrspace(5) undef
+ %tmp = load <2 x i16>, ptr addrspace(5) poison
%tmp1 = insertelement <2 x i16> %tmp, i16 42, i64 0
- store <2 x i16> %tmp1, ptr addrspace(5) undef
+ store <2 x i16> %tmp1, ptr addrspace(5) poison
ret void
}
@@ -15,8 +15,8 @@ define void @scalar_to_vector_i16() {
; GCN: v_mov_b32_e32 [[V:v[0-9]+]], 0x3c00
; GCN: buffer_store_short [[V]],
define void @scalar_to_vector_f16() {
- %tmp = load <2 x half>, ptr addrspace(5) undef
+ %tmp = load <2 x half>, ptr addrspace(5) poison
%tmp1 = insertelement <2 x half> %tmp, half 1.0, i64 0
- store <2 x half> %tmp1, ptr addrspace(5) undef
+ store <2 x half> %tmp1, ptr addrspace(5) poison
ret void
}
diff --git
a/llvm/test/CodeGen/AMDGPU/stack-pointer-offset-relative-frameindex.ll
b/llvm/test/CodeGen/AMDGPU/stack-pointer-offset-relative-frameindex.ll
index d4d3b37a0ed1e..dcf0d3d1829cd 100644
--- a/llvm/test/CodeGen/AMDGPU/stack-pointer-offset-relative-frameindex.ll
+++ b/llvm/test/CodeGen/AMDGPU/stack-pointer-offset-relative-frameindex.ll
@@ -148,7 +148,7 @@ if.then4.i: ; preds =
%entry
%add1.i.i = add i32 %add.i.i, 0
%mul.i.i.i.i = mul i32 %add1.i.i, 1103515245
%add.i.i.i.i = add i32 %mul.i.i.i.i, 12345
- store i32 %add.i.i.i.i, ptr addrspace(5) undef, align 16
+ store i32 %add.i.i.i.i, ptr addrspace(5) poison, align 16
br label %shader_eval_surface.exit
shader_eval_surface.exit: ; preds = %entry
_______________________________________________
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
