[llvm] AMDGPU: Reorganize fmed3 intrinsic instcombine tests (PR #139498)
via llvm-commits
llvm-commits at lists.llvm.org
Sun May 11 22:58:20 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-amdgpu
Author: Matt Arsenault (arsenm)
<details>
<summary>Changes</summary>
---
Patch is 90.27 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/139498.diff
3 Files Affected:
- (modified) llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll (-294)
- (added) llvm/test/Transforms/InstCombine/AMDGPU/fmed3-fpext-fold.ll (+613)
- (modified) llvm/test/Transforms/InstCombine/AMDGPU/fmed3.ll (+213-529)
``````````diff
diff --git a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
index 6981b43244575..89c73c2161e35 100644
--- a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
@@ -1851,300 +1851,6 @@ define void @exp_compr_disabled_inputs_to_undef(<2 x half> %xy, <2 x half> %zw)
ret void
}
-; --------------------------------------------------------------------
-; llvm.amdgcn.fmed3
-; --------------------------------------------------------------------
-
-declare float @llvm.amdgcn.fmed3.f32(float, float, float) nounwind readnone
-
-define float @fmed3_f32(float %x, float %y, float %z) {
-; CHECK-LABEL: @fmed3_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[X:%.*]], float [[Y:%.*]], float [[Z:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float %y, float %z)
- ret float %med3
-}
-
-define float @fmed3_canonicalize_x_c0_c1_f32(float %x) {
-; CHECK-LABEL: @fmed3_canonicalize_x_c0_c1_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[X:%.*]], float 0.000000e+00, float 1.000000e+00)
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float 0.0, float 1.0)
- ret float %med3
-}
-
-define float @fmed3_canonicalize_c0_x_c1_f32(float %x) {
-; CHECK-LABEL: @fmed3_canonicalize_c0_x_c1_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[X:%.*]], float 0.000000e+00, float 1.000000e+00)
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0.0, float %x, float 1.0)
- ret float %med3
-}
-
-define float @fmed3_canonicalize_c0_c1_x_f32(float %x) {
-; CHECK-LABEL: @fmed3_canonicalize_c0_c1_x_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[X:%.*]], float 0.000000e+00, float 1.000000e+00)
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0.0, float 1.0, float %x)
- ret float %med3
-}
-
-define float @fmed3_canonicalize_x_y_c_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_canonicalize_x_y_c_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[X:%.*]], float [[Y:%.*]], float 1.000000e+00)
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float %y, float 1.0)
- ret float %med3
-}
-
-define float @fmed3_canonicalize_x_c_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_canonicalize_x_c_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[X:%.*]], float [[Y:%.*]], float 1.000000e+00)
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float 1.0, float %y)
- ret float %med3
-}
-
-define float @fmed3_canonicalize_c_x_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_canonicalize_c_x_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[X:%.*]], float [[Y:%.*]], float 1.000000e+00)
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 1.0, float %x, float %y)
- ret float %med3
-}
-
-define float @fmed3_undef_x_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_undef_x_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.minnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float undef, float %x, float %y)
- ret float %med3
-}
-
-define float @fmed3_fmf_undef_x_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_fmf_undef_x_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call nnan float @llvm.minnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call nnan float @llvm.amdgcn.fmed3.f32(float undef, float %x, float %y)
- ret float %med3
-}
-
-define float @fmed3_x_undef_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_x_undef_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.minnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float undef, float %y)
- ret float %med3
-}
-
-define float @fmed3_x_y_undef_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_x_y_undef_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.maxnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float %y, float undef)
- ret float %med3
-}
-
-define float @fmed3_qnan0_x_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_qnan0_x_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.minnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0x7FF8000000000000, float %x, float %y)
- ret float %med3
-}
-
-define float @fmed3_x_qnan0_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_x_qnan0_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.minnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float 0x7FF8000000000000, float %y)
- ret float %med3
-}
-
-define float @fmed3_x_y_qnan0_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_x_y_qnan0_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.maxnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float %y, float 0x7FF8000000000000)
- ret float %med3
-}
-
-define float @fmed3_qnan1_x_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_qnan1_x_y_f32(
-; CHECK-NEXT: [[MED3:%.*]] = call float @llvm.minnum.f32(float [[X:%.*]], float [[Y:%.*]])
-; CHECK-NEXT: ret float [[MED3]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0x7FF8000100000000, float %x, float %y)
- ret float %med3
-}
-
-; This can return any of the qnans.
-define float @fmed3_qnan0_qnan1_qnan2_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_qnan0_qnan1_qnan2_f32(
-; CHECK-NEXT: ret float 0x7FF8030000000000
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0x7FF8000100000000, float 0x7FF8002000000000, float 0x7FF8030000000000)
- ret float %med3
-}
-
-define float @fmed3_constant_src0_0_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_constant_src0_0_f32(
-; CHECK-NEXT: ret float 5.000000e-01
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0.5, float -1.0, float 4.0)
- ret float %med3
-}
-
-define float @fmed3_constant_src0_1_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_constant_src0_1_f32(
-; CHECK-NEXT: ret float 5.000000e-01
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0.5, float 4.0, float -1.0)
- ret float %med3
-}
-
-define float @fmed3_constant_src1_0_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_constant_src1_0_f32(
-; CHECK-NEXT: ret float 5.000000e-01
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float -1.0, float 0.5, float 4.0)
- ret float %med3
-}
-
-define float @fmed3_constant_src1_1_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_constant_src1_1_f32(
-; CHECK-NEXT: ret float 5.000000e-01
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 4.0, float 0.5, float -1.0)
- ret float %med3
-}
-
-define float @fmed3_constant_src2_0_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_constant_src2_0_f32(
-; CHECK-NEXT: ret float 5.000000e-01
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float -1.0, float 4.0, float 0.5)
- ret float %med3
-}
-
-define float @fmed3_constant_src2_1_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_constant_src2_1_f32(
-; CHECK-NEXT: ret float 5.000000e-01
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 4.0, float -1.0, float 0.5)
- ret float %med3
-}
-
-define float @fmed3_x_qnan0_qnan1_f32(float %x) {
-; CHECK-LABEL: @fmed3_x_qnan0_qnan1_f32(
-; CHECK-NEXT: ret float [[X:%.*]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float 0x7FF8001000000000, float 0x7FF8002000000000)
- ret float %med3
-}
-
-define float @fmed3_qnan0_x_qnan1_f32(float %x) {
-; CHECK-LABEL: @fmed3_qnan0_x_qnan1_f32(
-; CHECK-NEXT: ret float [[X:%.*]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0x7FF8001000000000, float %x, float 0x7FF8002000000000)
- ret float %med3
-}
-
-define float @fmed3_qnan0_qnan1_x_f32(float %x) {
-; CHECK-LABEL: @fmed3_qnan0_qnan1_x_f32(
-; CHECK-NEXT: ret float [[X:%.*]]
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0x7FF8001000000000, float 0x7FF8002000000000, float %x)
- ret float %med3
-}
-
-define float @fmed3_nan_0_1_f32() {
-; CHECK-LABEL: @fmed3_nan_0_1_f32(
-; CHECK-NEXT: ret float 0.000000e+00
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float 0x7FF8001000000000, float 0.0, float 1.0)
- ret float %med3
-}
-
-define float @fmed3_0_nan_1_f32() {
-; CHECK-LABEL: @fmed3_0_nan_1_f32(
-; CHECK-NEXT: ret float 0.000000e+00
-;
- %med = call float @llvm.amdgcn.fmed3.f32(float 0.0, float 0x7FF8001000000000, float 1.0)
- ret float %med
-}
-
-define float @fmed3_0_1_nan_f32() {
-; CHECK-LABEL: @fmed3_0_1_nan_f32(
-; CHECK-NEXT: ret float 1.000000e+00
-;
- %med = call float @llvm.amdgcn.fmed3.f32(float 0.0, float 1.0, float 0x7FF8001000000000)
- ret float %med
-}
-
-define float @fmed3_undef_0_1_f32() {
-; CHECK-LABEL: @fmed3_undef_0_1_f32(
-; CHECK-NEXT: ret float 0.000000e+00
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float undef, float 0.0, float 1.0)
- ret float %med3
-}
-
-define float @fmed3_0_undef_1_f32() {
-; CHECK-LABEL: @fmed3_0_undef_1_f32(
-; CHECK-NEXT: ret float 0.000000e+00
-;
- %med = call float @llvm.amdgcn.fmed3.f32(float 0.0, float undef, float 1.0)
- ret float %med
-}
-
-define float @fmed3_0_1_undef_f32() {
-; CHECK-LABEL: @fmed3_0_1_undef_f32(
-; CHECK-NEXT: ret float 1.000000e+00
-;
- %med = call float @llvm.amdgcn.fmed3.f32(float 0.0, float 1.0, float undef)
- ret float %med
-}
-
-define float @fmed3_poison_x_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_poison_x_y_f32(
-; CHECK-NEXT: ret float poison
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float poison, float %x, float %y)
- ret float %med3
-}
-
-define float @fmed3_x_poison_y_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_x_poison_y_f32(
-; CHECK-NEXT: ret float poison
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float poison, float %y)
- ret float %med3
-}
-
-define float @fmed3_x_y_poison_f32(float %x, float %y) {
-; CHECK-LABEL: @fmed3_x_y_poison_f32(
-; CHECK-NEXT: ret float poison
-;
- %med3 = call float @llvm.amdgcn.fmed3.f32(float %x, float %y, float poison)
- ret float %med3
-}
-
; --------------------------------------------------------------------
; llvm.amdgcn.icmp
; --------------------------------------------------------------------
diff --git a/llvm/test/Transforms/InstCombine/AMDGPU/fmed3-fpext-fold.ll b/llvm/test/Transforms/InstCombine/AMDGPU/fmed3-fpext-fold.ll
new file mode 100644
index 0000000000000..a31b47b2ca6e7
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/AMDGPU/fmed3-fpext-fold.ll
@@ -0,0 +1,613 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
+; Unknown/default target
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=instcombine < %s | FileCheck -check-prefixes=NO-FMED3F16,UNKNOWN %s
+
+; Known target, no med3_f16
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -passes=instcombine < %s | FileCheck -check-prefixes=NO-FMED3F16,GFX8 %s
+
+; Has med3_f16
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -passes=instcombine < %s | FileCheck -check-prefixes=GFX9 %s
+
+
+declare float @llvm.fabs.f32(float) #0
+declare half @llvm.fabs.f16(half) #0
+declare float @llvm.amdgcn.fmed3.f32(float, float, float) #0
+
+define float @fmed3_f32_fpext_f16(half %arg0, half %arg1, half %arg2) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16
+; NO-FMED3F16-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1:[0-9]+]] {
+; NO-FMED3F16-NEXT: [[ARG0_EXT:%.*]] = fpext half [[ARG0]] to float
+; NO-FMED3F16-NEXT: [[ARG1_EXT:%.*]] = fpext half [[ARG1]] to float
+; NO-FMED3F16-NEXT: [[ARG2_EXT:%.*]] = fpext half [[ARG2]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[ARG0_EXT]], float [[ARG1_EXT]], float [[ARG2_EXT]])
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16
+; GFX9-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1:[0-9]+]] {
+; GFX9-NEXT: [[MED31:%.*]] = call half @llvm.amdgcn.fmed3.f16(half [[ARG0]], half [[ARG1]], half [[ARG2]])
+; GFX9-NEXT: [[MED3:%.*]] = fpext half [[MED31]] to float
+; GFX9-NEXT: ret float [[MED3]]
+;
+ %arg0.ext = fpext half %arg0 to float
+ %arg1.ext = fpext half %arg1 to float
+ %arg2.ext = fpext half %arg2 to float
+ %med3 = call float @llvm.amdgcn.fmed3.f32(float %arg0.ext, float %arg1.ext, float %arg2.ext)
+ ret float %med3
+}
+
+define float @fmed3_f32_fpext_f16_flags(half %arg0, half %arg1, half %arg2) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16_flags
+; NO-FMED3F16-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; NO-FMED3F16-NEXT: [[ARG0_EXT:%.*]] = fpext half [[ARG0]] to float
+; NO-FMED3F16-NEXT: [[ARG1_EXT:%.*]] = fpext half [[ARG1]] to float
+; NO-FMED3F16-NEXT: [[ARG2_EXT:%.*]] = fpext half [[ARG2]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call nsz float @llvm.amdgcn.fmed3.f32(float [[ARG0_EXT]], float [[ARG1_EXT]], float [[ARG2_EXT]])
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16_flags
+; GFX9-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; GFX9-NEXT: [[MED31:%.*]] = call nsz half @llvm.amdgcn.fmed3.f16(half [[ARG0]], half [[ARG1]], half [[ARG2]])
+; GFX9-NEXT: [[MED3:%.*]] = fpext half [[MED31]] to float
+; GFX9-NEXT: ret float [[MED3]]
+;
+ %arg0.ext = fpext half %arg0 to float
+ %arg1.ext = fpext half %arg1 to float
+ %arg2.ext = fpext half %arg2 to float
+ %med3 = call nsz float @llvm.amdgcn.fmed3.f32(float %arg0.ext, float %arg1.ext, float %arg2.ext)
+ ret float %med3
+}
+
+define float @fmed3_f32_fpext_f16_k0(half %arg1, half %arg2) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16_k0
+; NO-FMED3F16-SAME: (half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; NO-FMED3F16-NEXT: [[ARG1_EXT:%.*]] = fpext half [[ARG1]] to float
+; NO-FMED3F16-NEXT: [[ARG2_EXT:%.*]] = fpext half [[ARG2]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[ARG1_EXT]], float [[ARG2_EXT]], float 2.000000e+00)
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16_k0
+; GFX9-SAME: (half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; GFX9-NEXT: [[MED31:%.*]] = call half @llvm.amdgcn.fmed3.f16(half [[ARG1]], half [[ARG2]], half 0xH4000)
+; GFX9-NEXT: [[MED3:%.*]] = fpext half [[MED31]] to float
+; GFX9-NEXT: ret float [[MED3]]
+;
+ %arg1.ext = fpext half %arg1 to float
+ %arg2.ext = fpext half %arg2 to float
+ %med3 = call float @llvm.amdgcn.fmed3.f32(float 2.0, float %arg1.ext, float %arg2.ext)
+ ret float %med3
+}
+
+define float @fmed3_f32_fpext_f16_k1(half %arg0, half %arg2) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16_k1
+; NO-FMED3F16-SAME: (half [[ARG0:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; NO-FMED3F16-NEXT: [[ARG0_EXT:%.*]] = fpext half [[ARG0]] to float
+; NO-FMED3F16-NEXT: [[ARG2_EXT:%.*]] = fpext half [[ARG2]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[ARG0_EXT]], float [[ARG2_EXT]], float 2.000000e+00)
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16_k1
+; GFX9-SAME: (half [[ARG0:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; GFX9-NEXT: [[MED31:%.*]] = call half @llvm.amdgcn.fmed3.f16(half [[ARG0]], half [[ARG2]], half 0xH4000)
+; GFX9-NEXT: [[MED3:%.*]] = fpext half [[MED31]] to float
+; GFX9-NEXT: ret float [[MED3]]
+;
+ %arg0.ext = fpext half %arg0 to float
+ %arg2.ext = fpext half %arg2 to float
+ %med3 = call float @llvm.amdgcn.fmed3.f32(float %arg0.ext, float 2.0, float %arg2.ext)
+ ret float %med3
+}
+
+define float @fmed3_f32_fpext_f16_k2(half %arg0, half %arg1) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16_k2
+; NO-FMED3F16-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]]) #[[ATTR1]] {
+; NO-FMED3F16-NEXT: [[ARG0_EXT:%.*]] = fpext half [[ARG0]] to float
+; NO-FMED3F16-NEXT: [[ARG1_EXT:%.*]] = fpext half [[ARG1]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[ARG0_EXT]], float [[ARG1_EXT]], float 2.000000e+00)
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16_k2
+; GFX9-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]]) #[[ATTR1]] {
+; GFX9-NEXT: [[MED31:%.*]] = call half @llvm.amdgcn.fmed3.f16(half [[ARG0]], half [[ARG1]], half 0xH4000)
+; GFX9-NEXT: [[MED3:%.*]] = fpext half [[MED31]] to float
+; GFX9-NEXT: ret float [[MED3]]
+;
+ %arg0.ext = fpext half %arg0 to float
+ %arg1.ext = fpext half %arg1 to float
+ %med3 = call float @llvm.amdgcn.fmed3.f32(float %arg0.ext, float %arg1.ext, float 2.0)
+ ret float %med3
+}
+
+define float @fmed3_f32_fpext_f16_k0_k1(half %arg2) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16_k0_k1
+; NO-FMED3F16-SAME: (half [[ARG2:%.*]]) #[[ATTR1]] {
+; NO-FMED3F16-NEXT: [[ARG2_EXT:%.*]] = fpext half [[ARG2]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[ARG2_EXT]], float 0.000000e+00, float 1.600000e+01)
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16_k0_k1
+; GFX9-SAME: (half [[ARG2:%.*]]) #[[ATTR1]] {
+; GFX9-NEXT: [[MED31:%.*]] = call half @llvm.amdgcn.fmed3.f16(half [[ARG2]], half 0xH0000, half 0xH4C00)
+; GFX9-NEXT: [[MED3:%.*]] = fpext half [[MED31]] to float
+; GFX9-NEXT: ret float [[MED3]]
+;
+ %arg2.ext = fpext half %arg2 to float
+ %med3 = call float @llvm.amdgcn.fmed3.f32(float 0.0, float 16.0, float %arg2.ext)
+ ret float %med3
+}
+
+define float @fmed3_f32_fpext_f16_k0_k2(half %arg1) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16_k0_k2
+; NO-FMED3F16-SAME: (half [[ARG1:%.*]]) #[[ATTR1]] {
+; NO-FMED3F16-NEXT: [[ARG1_EXT:%.*]] = fpext half [[ARG1]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[ARG1_EXT]], float 0.000000e+00, float 2.000000e+00)
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16_k0_k2
+; GFX9-SAME: (half [[ARG1:%.*]]) #[[ATTR1]] {
+; GFX9-NEXT: [[MED31:%.*]] = call half @llvm.amdgcn.fmed3.f16(half [[ARG1]], half 0xH0000, half 0xH4000)
+; GFX9-NEXT: [[MED3:%.*]] = fpext half [[MED31]] to float
+; GFX9-NEXT: ret float [[MED3]]
+;
+ %arg1.ext = fpext half %arg1 to float
+ %med3 = call float @llvm.amdgcn.fmed3.f32(float 0.0, float %arg1.ext, float 2.0)
+ ret float %med3
+}
+
+define float @fmed3_f32_fpext_f16_fabs(half %arg0, half %arg1, half %arg2) #1 {
+; NO-FMED3F16-LABEL: define float @fmed3_f32_fpext_f16_fabs
+; NO-FMED3F16-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; NO-FMED3F16-NEXT: [[FABS_ARG0:%.*]] = call half @llvm.fabs.f16(half [[ARG0]])
+; NO-FMED3F16-NEXT: [[FABS_ARG1:%.*]] = call half @llvm.fabs.f16(half [[ARG1]])
+; NO-FMED3F16-NEXT: [[FABS_ARG2:%.*]] = call half @llvm.fabs.f16(half [[ARG2]])
+; NO-FMED3F16-NEXT: [[ARG0_EXT:%.*]] = fpext half [[FABS_ARG0]] to float
+; NO-FMED3F16-NEXT: [[ARG1_EXT:%.*]] = fpext half [[FABS_ARG1]] to float
+; NO-FMED3F16-NEXT: [[ARG2_EXT:%.*]] = fpext half [[FABS_ARG2]] to float
+; NO-FMED3F16-NEXT: [[MED3:%.*]] = call float @llvm.amdgcn.fmed3.f32(float [[ARG0_EXT]], float [[ARG1_EXT]], float [[ARG2_EXT]])
+; NO-FMED3F16-NEXT: ret float [[MED3]]
+;
+; GFX9-LABEL: define float @fmed3_f32_fpext_f16_fabs
+; GFX9-SAME: (half [[ARG0:%.*]], half [[ARG1:%.*]], half [[ARG2:%.*]]) #[[ATTR1]] {
+; GFX9-NEXT: [[FABS_ARG0:%.*]] = call half @llvm.fabs.f16(half [[ARG0]])
+; GFX9-NEXT: [[FABS_ARG1:%.*]] = call half @llvm.fabs.f16(half [[ARG1]])
+; GFX9-NEXT: [[FABS_ARG2:%.*]] = call half @llvm....
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/139498
More information about the llvm-commits
mailing list