aboutsummaryrefslogtreecommitdiff
path: root/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll
diff options
context:
space:
mode:
Diffstat (limited to 'test/CodeGen/AMDGPU/llvm.amdgcn.class.ll')
-rw-r--r--test/CodeGen/AMDGPU/llvm.amdgcn.class.ll60
1 files changed, 30 insertions, 30 deletions
diff --git a/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll b/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll
index 668c669e41e8..1fcdac537fba 100644
--- a/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll
+++ b/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll
@@ -14,7 +14,7 @@ declare double @llvm.fabs.f64(double) #1
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
%result = call i1 @llvm.amdgcn.class.f32(float %a, i32 %b) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -29,7 +29,7 @@ define void @test_class_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[CMP]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_fabs_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_fabs_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
%a.fabs = call float @llvm.fabs.f32(float %a) #1
%result = call i1 @llvm.amdgcn.class.f32(float %a.fabs, i32 %b) #1
%sext = sext i1 %result to i32
@@ -45,7 +45,7 @@ define void @test_class_fabs_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[CMP]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_fneg_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_fneg_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
%a.fneg = fsub float -0.0, %a
%result = call i1 @llvm.amdgcn.class.f32(float %a.fneg, i32 %b) #1
%sext = sext i1 %result to i32
@@ -61,7 +61,7 @@ define void @test_class_fneg_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[CMP]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_fneg_fabs_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_fneg_fabs_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
%a.fabs = call float @llvm.fabs.f32(float %a) #1
%a.fneg.fabs = fsub float -0.0, %a.fabs
%result = call i1 @llvm.amdgcn.class.f32(float %a.fneg.fabs, i32 %b) #1
@@ -76,7 +76,7 @@ define void @test_class_fneg_fabs_f32(i32 addrspace(1)* %out, float %a, i32 %b)
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[COND]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_1_f32(i32 addrspace(1)* %out, float %a) #0 {
+define amdgpu_kernel void @test_class_1_f32(i32 addrspace(1)* %out, float %a) #0 {
%result = call i1 @llvm.amdgcn.class.f32(float %a, i32 1) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -89,7 +89,7 @@ define void @test_class_1_f32(i32 addrspace(1)* %out, float %a) #0 {
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[COND]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_64_f32(i32 addrspace(1)* %out, float %a) #0 {
+define amdgpu_kernel void @test_class_64_f32(i32 addrspace(1)* %out, float %a) #0 {
%result = call i1 @llvm.amdgcn.class.f32(float %a, i32 64) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -104,7 +104,7 @@ define void @test_class_64_f32(i32 addrspace(1)* %out, float %a) #0 {
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_full_mask_f32(i32 addrspace(1)* %out, float %a) #0 {
+define amdgpu_kernel void @test_class_full_mask_f32(i32 addrspace(1)* %out, float %a) #0 {
%result = call i1 @llvm.amdgcn.class.f32(float %a, i32 1023) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -118,7 +118,7 @@ define void @test_class_full_mask_f32(i32 addrspace(1)* %out, float %a) #0 {
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_9bit_mask_f32(i32 addrspace(1)* %out, float %a) #0 {
+define amdgpu_kernel void @test_class_9bit_mask_f32(i32 addrspace(1)* %out, float %a) #0 {
%result = call i1 @llvm.amdgcn.class.f32(float %a, i32 511) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -132,7 +132,7 @@ define void @test_class_9bit_mask_f32(i32 addrspace(1)* %out, float %a) #0 {
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @v_test_class_full_mask_f32(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
+define amdgpu_kernel void @v_test_class_full_mask_f32(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr float, float addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -150,7 +150,7 @@ define void @v_test_class_full_mask_f32(i32 addrspace(1)* %out, float addrspace(
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_inline_imm_constant_dynamic_mask_f32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_class_inline_imm_constant_dynamic_mask_f32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -170,7 +170,7 @@ define void @test_class_inline_imm_constant_dynamic_mask_f32(i32 addrspace(1)* %
; SI-NEXT: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_lit_constant_dynamic_mask_f32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_class_lit_constant_dynamic_mask_f32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -190,7 +190,7 @@ define void @test_class_lit_constant_dynamic_mask_f32(i32 addrspace(1)* %out, i3
; SI: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
%result = call i1 @llvm.amdgcn.class.f64(double %a, i32 %b) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -205,7 +205,7 @@ define void @test_class_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
; SI: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[CMP]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_fabs_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_fabs_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
%a.fabs = call double @llvm.fabs.f64(double %a) #1
%result = call i1 @llvm.amdgcn.class.f64(double %a.fabs, i32 %b) #1
%sext = sext i1 %result to i32
@@ -221,7 +221,7 @@ define void @test_class_fabs_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
; SI: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[CMP]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_fneg_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_fneg_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
%a.fneg = fsub double -0.0, %a
%result = call i1 @llvm.amdgcn.class.f64(double %a.fneg, i32 %b) #1
%sext = sext i1 %result to i32
@@ -237,7 +237,7 @@ define void @test_class_fneg_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
; SI: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, [[CMP]]
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_fneg_fabs_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_fneg_fabs_f64(i32 addrspace(1)* %out, double %a, i32 %b) #0 {
%a.fabs = call double @llvm.fabs.f64(double %a) #1
%a.fneg.fabs = fsub double -0.0, %a.fabs
%result = call i1 @llvm.amdgcn.class.f64(double %a.fneg.fabs, i32 %b) #1
@@ -249,7 +249,7 @@ define void @test_class_fneg_fabs_f64(i32 addrspace(1)* %out, double %a, i32 %b)
; SI-LABEL: {{^}}test_class_1_f64:
; SI: v_cmp_class_f64_e64 {{s\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 1{{$}}
; SI: s_endpgm
-define void @test_class_1_f64(i32 addrspace(1)* %out, double %a) #0 {
+define amdgpu_kernel void @test_class_1_f64(i32 addrspace(1)* %out, double %a) #0 {
%result = call i1 @llvm.amdgcn.class.f64(double %a, i32 1) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -259,7 +259,7 @@ define void @test_class_1_f64(i32 addrspace(1)* %out, double %a) #0 {
; SI-LABEL: {{^}}test_class_64_f64:
; SI: v_cmp_class_f64_e64 {{s\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 64{{$}}
; SI: s_endpgm
-define void @test_class_64_f64(i32 addrspace(1)* %out, double %a) #0 {
+define amdgpu_kernel void @test_class_64_f64(i32 addrspace(1)* %out, double %a) #0 {
%result = call i1 @llvm.amdgcn.class.f64(double %a, i32 64) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -275,7 +275,7 @@ define void @test_class_64_f64(i32 addrspace(1)* %out, double %a) #0 {
; SI: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI-NEXT: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_full_mask_f64(i32 addrspace(1)* %out, double %a) #0 {
+define amdgpu_kernel void @test_class_full_mask_f64(i32 addrspace(1)* %out, double %a) #0 {
%result = call i1 @llvm.amdgcn.class.f64(double %a, i32 511) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -290,7 +290,7 @@ define void @test_class_full_mask_f64(i32 addrspace(1)* %out, double %a) #0 {
; SI: v_cndmask_b32_e64 [[RESULT:v[0-9]+]], 0, -1, vcc
; SI: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @v_test_class_full_mask_f64(i32 addrspace(1)* %out, double addrspace(1)* %in) #0 {
+define amdgpu_kernel void @v_test_class_full_mask_f64(i32 addrspace(1)* %out, double addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr double, double addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -306,7 +306,7 @@ define void @v_test_class_full_mask_f64(i32 addrspace(1)* %out, double addrspace
; XSI: v_cmp_class_f64_e32 vcc, 1.0,
; SI: v_cmp_class_f64_e32 vcc,
; SI: s_endpgm
-define void @test_class_inline_imm_constant_dynamic_mask_f64(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_class_inline_imm_constant_dynamic_mask_f64(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -321,7 +321,7 @@ define void @test_class_inline_imm_constant_dynamic_mask_f64(i32 addrspace(1)* %
; SI-LABEL: {{^}}test_class_lit_constant_dynamic_mask_f64:
; SI: v_cmp_class_f64_e32 vcc, s{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}
; SI: s_endpgm
-define void @test_class_lit_constant_dynamic_mask_f64(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_class_lit_constant_dynamic_mask_f64(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -338,7 +338,7 @@ define void @test_class_lit_constant_dynamic_mask_f64(i32 addrspace(1)* %out, i3
; SI: v_cmp_class_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, 3{{$}}
; SI-NOT: v_cmp_class
; SI: s_endpgm
-define void @test_fold_or_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fold_or_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr float, float addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -358,7 +358,7 @@ define void @test_fold_or_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)
; SI: v_cmp_class_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, 7{{$}}
; SI-NOT: v_cmp_class
; SI: s_endpgm
-define void @test_fold_or3_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fold_or3_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr float, float addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -381,7 +381,7 @@ define void @test_fold_or3_class_f32_0(i32 addrspace(1)* %out, float addrspace(1
; SI: v_cmp_class_f32_e32 vcc, v{{[0-9]+}}, [[MASK]]{{$}}
; SI-NOT: v_cmp_class
; SI: s_endpgm
-define void @test_fold_or_all_tests_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fold_or_all_tests_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr float, float addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -416,7 +416,7 @@ define void @test_fold_or_all_tests_class_f32_0(i32 addrspace(1)* %out, float ad
; SI: v_cmp_class_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, 12{{$}}
; SI-NOT: v_cmp_class
; SI: s_endpgm
-define void @test_fold_or_class_f32_1(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fold_or_class_f32_1(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr float, float addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -436,7 +436,7 @@ define void @test_fold_or_class_f32_1(i32 addrspace(1)* %out, float addrspace(1)
; SI: v_cmp_class_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, 7{{$}}
; SI-NOT: v_cmp_class
; SI: s_endpgm
-define void @test_fold_or_class_f32_2(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fold_or_class_f32_2(i32 addrspace(1)* %out, float addrspace(1)* %in) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr float, float addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -456,7 +456,7 @@ define void @test_fold_or_class_f32_2(i32 addrspace(1)* %out, float addrspace(1)
; SI-DAG: v_cmp_class_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}}, 8{{$}}
; SI: s_or_b64
; SI: s_endpgm
-define void @test_no_fold_or_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in, float %b) #0 {
+define amdgpu_kernel void @test_no_fold_or_class_f32_0(i32 addrspace(1)* %out, float addrspace(1)* %in, float %b) #0 {
%tid = call i32 @llvm.amdgcn.workitem.id.x() #1
%gep.in = getelementptr float, float addrspace(1)* %in, i32 %tid
%gep.out = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
@@ -476,7 +476,7 @@ define void @test_no_fold_or_class_f32_0(i32 addrspace(1)* %out, float addrspace
; SI: v_mov_b32_e32 [[RESULT:v[0-9]+]], 0{{$}}
; SI: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_0_f32(i32 addrspace(1)* %out, float %a) #0 {
+define amdgpu_kernel void @test_class_0_f32(i32 addrspace(1)* %out, float %a) #0 {
%result = call i1 @llvm.amdgcn.class.f32(float %a, i32 0) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -488,7 +488,7 @@ define void @test_class_0_f32(i32 addrspace(1)* %out, float %a) #0 {
; SI: v_mov_b32_e32 [[RESULT:v[0-9]+]], 0{{$}}
; SI: buffer_store_dword [[RESULT]]
; SI: s_endpgm
-define void @test_class_0_f64(i32 addrspace(1)* %out, double %a) #0 {
+define amdgpu_kernel void @test_class_0_f64(i32 addrspace(1)* %out, double %a) #0 {
%result = call i1 @llvm.amdgcn.class.f64(double %a, i32 0) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
@@ -500,7 +500,7 @@ define void @test_class_0_f64(i32 addrspace(1)* %out, double %a) #0 {
; SI-NOT: v_cmp_class
; SI: v_cndmask_b32_e64 v{{[0-9]+}}, 0, -1,
; SI: buffer_store_dword
-define void @test_class_undef_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
+define amdgpu_kernel void @test_class_undef_f32(i32 addrspace(1)* %out, float %a, i32 %b) #0 {
%result = call i1 @llvm.amdgcn.class.f32(float undef, i32 %b) #1
%sext = sext i1 %result to i32
store i32 %sext, i32 addrspace(1)* %out, align 4
generated by cgit v1.2.3 (git 2.25.1) at 2025-03-09 20:01:27 +0000