[llvm] [AMDGPU][test] Remove unsafe-fp-math uses (NFC) (PR #164609)
via llvm-commits
llvm-commits at lists.llvm.org
Wed Oct 22 18:17:47 PDT 2025
https://github.com/paperchalice updated https://github.com/llvm/llvm-project/pull/164609
>From d65cb03c7b6b912d6218d6a81eda834a54995473 Mon Sep 17 00:00:00 2001
From: PaperChalice <liujunchang97 at outlook.com>
Date: Wed, 22 Oct 2025 19:59:25 +0800
Subject: [PATCH 1/2] [AMDGPU][test] Remove unsafe-fp-math uses (NFC)
---
.../CodeGen/AMDGPU/combine_andor_with_cmps.ll | 2 +-
llvm/test/CodeGen/AMDGPU/fdiv.f64.ll | 7 +++----
...ad-formation-fmul-distribute-denormal-mode.ll | 2 +-
llvm/test/CodeGen/AMDGPU/fmed3.bf16.ll | 4 ++--
llvm/test/CodeGen/AMDGPU/fmed3.ll | 2 +-
.../CodeGen/AMDGPU/fneg-combines.legal.f16.ll | 2 +-
llvm/test/CodeGen/AMDGPU/fneg-combines.ll | 2 +-
llvm/test/CodeGen/AMDGPU/frem.ll | 5 +----
llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll | 3 +--
llvm/test/CodeGen/AMDGPU/fsqrt.r600.ll | 2 +-
llvm/test/CodeGen/AMDGPU/inline-attr.ll | 16 ++++++++--------
llvm/test/CodeGen/AMDGPU/llvm.exp2.ll | 2 +-
llvm/test/CodeGen/AMDGPU/llvm.log2.ll | 2 +-
llvm/test/CodeGen/AMDGPU/minmax.ll | 2 +-
14 files changed, 24 insertions(+), 29 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll b/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll
index 57a1e4cb795bf..ec92edbe2bf65 100644
--- a/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll
+++ b/llvm/test/CodeGen/AMDGPU/combine_andor_with_cmps.ll
@@ -3385,7 +3385,7 @@ declare half @llvm.canonicalize.f16(half)
declare <2 x half> @llvm.canonicalize.v2f16(<2 x half>)
attributes #0 = { nounwind "amdgpu-ieee"="false" }
-attributes #1 = { nounwind "unsafe-fp-math"="true" "no-nans-fp-math"="true" }
+attributes #1 = { nounwind "no-nans-fp-math"="true" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX11NONANS-FAKE16: {{.*}}
; GFX11NONANS-TRUE16: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/fdiv.f64.ll b/llvm/test/CodeGen/AMDGPU/fdiv.f64.ll
index acb32d4951c1c..11476a6aa862e 100644
--- a/llvm/test/CodeGen/AMDGPU/fdiv.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fdiv.f64.ll
@@ -127,7 +127,7 @@ define amdgpu_kernel void @s_fdiv_v4f64(ptr addrspace(1) %out, <4 x double> %num
; GCN-LABEL: {{^}}div_fast_2_x_pat_f64:
; GCN: v_mul_f64 [[MUL:v\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, 0.5
; GCN: buffer_store_dwordx2 [[MUL]]
-define amdgpu_kernel void @div_fast_2_x_pat_f64(ptr addrspace(1) %out) #1 {
+define amdgpu_kernel void @div_fast_2_x_pat_f64(ptr addrspace(1) %out) #0 {
%x = load double, ptr addrspace(1) poison
%rcp = fdiv fast double %x, 2.0
store double %rcp, ptr addrspace(1) %out, align 4
@@ -139,7 +139,7 @@ define amdgpu_kernel void @div_fast_2_x_pat_f64(ptr addrspace(1) %out) #1 {
; GCN-DAG: v_mov_b32_e32 v[[K_HI:[0-9]+]], 0x3fb99999
; GCN: v_mul_f64 [[MUL:v\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, v[[[K_LO]]:[[K_HI]]]
; GCN: buffer_store_dwordx2 [[MUL]]
-define amdgpu_kernel void @div_fast_k_x_pat_f64(ptr addrspace(1) %out) #1 {
+define amdgpu_kernel void @div_fast_k_x_pat_f64(ptr addrspace(1) %out) #0 {
%x = load double, ptr addrspace(1) poison
%rcp = fdiv fast double %x, 10.0
store double %rcp, ptr addrspace(1) %out, align 4
@@ -151,7 +151,7 @@ define amdgpu_kernel void @div_fast_k_x_pat_f64(ptr addrspace(1) %out) #1 {
; GCN-DAG: v_mov_b32_e32 v[[K_HI:[0-9]+]], 0xbfb99999
; GCN: v_mul_f64 [[MUL:v\[[0-9]+:[0-9]+\]]], s{{\[[0-9]+:[0-9]+\]}}, v[[[K_LO]]:[[K_HI]]]
; GCN: buffer_store_dwordx2 [[MUL]]
-define amdgpu_kernel void @div_fast_neg_k_x_pat_f64(ptr addrspace(1) %out) #1 {
+define amdgpu_kernel void @div_fast_neg_k_x_pat_f64(ptr addrspace(1) %out) #0 {
%x = load double, ptr addrspace(1) poison
%rcp = fdiv fast double %x, -10.0
store double %rcp, ptr addrspace(1) %out, align 4
@@ -159,4 +159,3 @@ define amdgpu_kernel void @div_fast_neg_k_x_pat_f64(ptr addrspace(1) %out) #1 {
}
attributes #0 = { nounwind }
-attributes #1 = { nounwind "unsafe-fp-math"="true" }
diff --git a/llvm/test/CodeGen/AMDGPU/fmad-formation-fmul-distribute-denormal-mode.ll b/llvm/test/CodeGen/AMDGPU/fmad-formation-fmul-distribute-denormal-mode.ll
index 92eb4a642dc92..0a266bcbc48ae 100644
--- a/llvm/test/CodeGen/AMDGPU/fmad-formation-fmul-distribute-denormal-mode.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmad-formation-fmul-distribute-denormal-mode.ll
@@ -284,4 +284,4 @@ define <2 x float> @unsafe_fast_fmul_fsub_ditribute_post_legalize(float %arg0, <
ret <2 x float> %tmp1
}
-attributes #0 = { "no-infs-fp-math"="true" "unsafe-fp-math"="true" }
+attributes #0 = { "no-infs-fp-math"="true" }
diff --git a/llvm/test/CodeGen/AMDGPU/fmed3.bf16.ll b/llvm/test/CodeGen/AMDGPU/fmed3.bf16.ll
index bc85dc2f1e9e1..3e513de22caf3 100644
--- a/llvm/test/CodeGen/AMDGPU/fmed3.bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmed3.bf16.ll
@@ -219,8 +219,8 @@ define <2 x bfloat> @v_test_fmed3_r_i_i_v2bf16_minimumnum_maximumnum(<2 x bfloat
}
attributes #0 = { nounwind readnone }
-attributes #1 = { nounwind "unsafe-fp-math"="false" "no-nans-fp-math"="false" }
-attributes #2 = { nounwind "unsafe-fp-math"="false" "no-nans-fp-math"="true" }
+attributes #1 = { nounwind "no-nans-fp-math"="false" }
+attributes #2 = { nounwind "no-nans-fp-math"="true" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX11: {{.*}}
; GFX11-SDAG: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/fmed3.ll b/llvm/test/CodeGen/AMDGPU/fmed3.ll
index 3145a272ae48f..60ac0b943faf4 100644
--- a/llvm/test/CodeGen/AMDGPU/fmed3.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmed3.ll
@@ -8905,4 +8905,4 @@ declare half @llvm.minnum.f16(half, half) #0
declare half @llvm.maxnum.f16(half, half) #0
attributes #0 = { nounwind readnone }
-attributes #2 = { nounwind "unsafe-fp-math"="false" "no-nans-fp-math"="true" }
+attributes #2 = { nounwind "no-nans-fp-math"="true" }
diff --git a/llvm/test/CodeGen/AMDGPU/fneg-combines.legal.f16.ll b/llvm/test/CodeGen/AMDGPU/fneg-combines.legal.f16.ll
index d8bbda19453fa..69d1ee3f533a6 100644
--- a/llvm/test/CodeGen/AMDGPU/fneg-combines.legal.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fneg-combines.legal.f16.ll
@@ -159,7 +159,7 @@ declare half @llvm.amdgcn.interp.p2.f16(float, float, i32, i32, i1, i32) #0
attributes #0 = { nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind readnone }
-attributes #2 = { nounwind "unsafe-fp-math"="true" }
+attributes #2 = { nounwind }
attributes #3 = { nounwind "no-signed-zeros-fp-math"="true" }
attributes #4 = { nounwind "amdgpu-ieee"="false" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
diff --git a/llvm/test/CodeGen/AMDGPU/fneg-combines.ll b/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
index aaea4f76ea49b..b3202cbe30d0b 100644
--- a/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
+++ b/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
@@ -8006,7 +8006,7 @@ declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #0
attributes #0 = { nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind readnone }
-attributes #2 = { nounwind "unsafe-fp-math"="true" }
+attributes #2 = { nounwind }
attributes #3 = { nounwind "no-signed-zeros-fp-math"="true" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GCN-NSZ: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/frem.ll b/llvm/test/CodeGen/AMDGPU/frem.ll
index 6f91222b2f396..55eb7bc9cba69 100644
--- a/llvm/test/CodeGen/AMDGPU/frem.ll
+++ b/llvm/test/CodeGen/AMDGPU/frem.ll
@@ -18918,7 +18918,4 @@ define amdgpu_kernel void @frem_v2f64_const(ptr addrspace(1) %out) #0 {
-attributes #0 = { nounwind "unsafe-fp-math"="false" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
-attributes #1 = { nounwind "unsafe-fp-math"="true" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
-
-
+attributes #0 = { nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
diff --git a/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll b/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll
index 1b74ddfb883e8..9b979816b95b8 100644
--- a/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll
@@ -2870,7 +2870,7 @@ define double @v_sqrt_f64__enough_unsafe_attrs(double %x) #3 {
ret double %result
}
-define double @v_sqrt_f64__unsafe_attr(double %x) #4 {
+define double @v_sqrt_f64__unsafe_attr(double %x) {
; GFX6-SDAG-LABEL: v_sqrt_f64__unsafe_attr:
; GFX6-SDAG: ; %bb.0:
; GFX6-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
@@ -3449,7 +3449,6 @@ declare i32 @llvm.amdgcn.readfirstlane(i32) #1
attributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
attributes #1 = { convergent nounwind willreturn memory(none) }
attributes #3 = { "no-nans-fp-math"="true" "no-infs-fp-math"="true" }
-attributes #4 = { "unsafe-fp-math"="true" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX6: {{.*}}
; GFX8: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/fsqrt.r600.ll b/llvm/test/CodeGen/AMDGPU/fsqrt.r600.ll
index 9f19bcb638cd7..c93c077706046 100644
--- a/llvm/test/CodeGen/AMDGPU/fsqrt.r600.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsqrt.r600.ll
@@ -239,4 +239,4 @@ declare <2 x float> @llvm.sqrt.v2f32(<2 x float> %in) #0
declare <4 x float> @llvm.sqrt.v4f32(<4 x float> %in) #0
attributes #0 = { nounwind readnone }
-attributes #1 = { nounwind "unsafe-fp-math"="true" }
+attributes #1 = { nounwind }
diff --git a/llvm/test/CodeGen/AMDGPU/inline-attr.ll b/llvm/test/CodeGen/AMDGPU/inline-attr.ll
index 4e93ecac71622..c33b3344fd51c 100644
--- a/llvm/test/CodeGen/AMDGPU/inline-attr.ll
+++ b/llvm/test/CodeGen/AMDGPU/inline-attr.ll
@@ -36,18 +36,18 @@ entry:
ret void
}
-attributes #0 = { nounwind "uniform-work-group-size"="false" "unsafe-fp-math"="true"}
-attributes #1 = { nounwind "less-precise-fpmad"="true" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "unsafe-fp-math"="true" }
+attributes #0 = { nounwind "uniform-work-group-size"="false"}
+attributes #1 = { nounwind "less-precise-fpmad"="true" "no-infs-fp-math"="true" "no-nans-fp-math"="true" }
;.
-; UNSAFE: attributes #[[ATTR0]] = { nounwind "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
-; UNSAFE: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
+; UNSAFE: attributes #[[ATTR0]] = { nounwind "uniform-work-group-size"="false" }
+; UNSAFE: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "uniform-work-group-size"="false" }
;.
-; NONANS: attributes #[[ATTR0]] = { nounwind "no-nans-fp-math"="true" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
-; NONANS: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="true" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
+; NONANS: attributes #[[ATTR0]] = { nounwind "no-nans-fp-math"="true" "uniform-work-group-size"="false" }
+; NONANS: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="true" "uniform-work-group-size"="false" }
;.
-; NOINFS: attributes #[[ATTR0]] = { nounwind "no-infs-fp-math"="true" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
-; NOINFS: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="true" "no-nans-fp-math"="false" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
+; NOINFS: attributes #[[ATTR0]] = { nounwind "no-infs-fp-math"="true" "uniform-work-group-size"="false" }
+; NOINFS: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="true" "no-nans-fp-math"="false" "uniform-work-group-size"="false" }
;.
; UNSAFE: [[META0]] = !{}
;.
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.exp2.ll b/llvm/test/CodeGen/AMDGPU/llvm.exp2.ll
index 883db20a867b3..e30a58699fadb 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.exp2.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.exp2.ll
@@ -1485,7 +1485,7 @@ define float @v_exp2_f32_fast(float %in) {
ret float %result
}
-define float @v_exp2_f32_unsafe_math_attr(float %in) "unsafe-fp-math"="true" {
+define float @v_exp2_f32_unsafe_math_attr(float %in) {
; SI-SDAG-LABEL: v_exp2_f32_unsafe_math_attr:
; SI-SDAG: ; %bb.0:
; SI-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log2.ll b/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
index 0854134be6f46..61a777f8877bb 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log2.ll
@@ -1907,7 +1907,7 @@ define float @v_log2_f32_fast(float %in) {
ret float %result
}
-define float @v_log2_f32_unsafe_math_attr(float %in) "unsafe-fp-math"="true" {
+define float @v_log2_f32_unsafe_math_attr(float %in) {
; SI-SDAG-LABEL: v_log2_f32_unsafe_math_attr:
; SI-SDAG: ; %bb.0:
; SI-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/minmax.ll b/llvm/test/CodeGen/AMDGPU/minmax.ll
index d578d2e9720f0..60570bdbfc23b 100644
--- a/llvm/test/CodeGen/AMDGPU/minmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/minmax.ll
@@ -1296,4 +1296,4 @@ declare half @llvm.minnum.f16(half, half)
declare half @llvm.maxnum.f16(half, half)
declare float @llvm.minnum.f32(float, float)
declare float @llvm.maxnum.f32(float, float)
-attributes #0 = { nounwind "unsafe-fp-math"="false" "no-nans-fp-math"="true" }
+attributes #0 = { nounwind "no-nans-fp-math"="true" }
>From f83532c1d94a48ab487a2985ed4df21f3bec9495 Mon Sep 17 00:00:00 2001
From: paperchalice <liujunchang97 at outlook.com>
Date: Thu, 23 Oct 2025 09:17:38 +0800
Subject: [PATCH 2/2] Fix attributes
---
llvm/test/CodeGen/AMDGPU/frem.ll | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/frem.ll b/llvm/test/CodeGen/AMDGPU/frem.ll
index 55eb7bc9cba69..d8cbdb11a911f 100644
--- a/llvm/test/CodeGen/AMDGPU/frem.ll
+++ b/llvm/test/CodeGen/AMDGPU/frem.ll
@@ -2048,7 +2048,7 @@ define amdgpu_kernel void @unsafe_frem_f16(ptr addrspace(1) %out, ptr addrspace(
; GFX1200-FAKE16-NEXT: v_fmac_f16_e32 v1, v3, v2
; GFX1200-FAKE16-NEXT: global_store_b16 v0, v1, s[0:1]
; GFX1200-FAKE16-NEXT: s_endpgm
- ptr addrspace(1) %in2) #1 {
+ ptr addrspace(1) %in2) #0 {
%gep2 = getelementptr half, ptr addrspace(1) %in2, i32 4
%r0 = load half, ptr addrspace(1) %in1, align 4
%r1 = load half, ptr addrspace(1) %gep2, align 4
@@ -3417,7 +3417,7 @@ define amdgpu_kernel void @unsafe_frem_f32(ptr addrspace(1) %out, ptr addrspace(
; GFX1200-NEXT: v_fmac_f32_e32 v1, v3, v2
; GFX1200-NEXT: global_store_b32 v0, v1, s[0:1]
; GFX1200-NEXT: s_endpgm
- ptr addrspace(1) %in2) #1 {
+ ptr addrspace(1) %in2) #0 {
%gep2 = getelementptr float, ptr addrspace(1) %in2, i32 4
%r0 = load float, ptr addrspace(1) %in1, align 4
%r1 = load float, ptr addrspace(1) %gep2, align 4
@@ -4821,7 +4821,7 @@ define amdgpu_kernel void @unsafe_frem_f64(ptr addrspace(1) %out, ptr addrspace(
; GFX1200-NEXT: v_fma_f64 v[0:1], -v[4:5], v[2:3], v[0:1]
; GFX1200-NEXT: global_store_b64 v12, v[0:1], s[0:1]
; GFX1200-NEXT: s_endpgm
- ptr addrspace(1) %in2) #1 {
+ ptr addrspace(1) %in2) #0 {
%r0 = load double, ptr addrspace(1) %in1, align 8
%r1 = load double, ptr addrspace(1) %in2, align 8
%r2 = frem afn double %r0, %r1
More information about the llvm-commits
mailing list