diff options
Diffstat (limited to 'clang/test/CodeGen/builtins-image-load.c')
| -rw-r--r-- | clang/test/CodeGen/builtins-image-load.c | 1203 |
1 files changed, 1203 insertions, 0 deletions
diff --git a/clang/test/CodeGen/builtins-image-load.c b/clang/test/CodeGen/builtins-image-load.c new file mode 100644 index 0000000..8442124 --- /dev/null +++ b/clang/test/CodeGen/builtins-image-load.c @@ -0,0 +1,1203 @@ +// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5 +// RUN: %clang_cc1 -triple amdgcn-- -target-cpu gfx1100 %s -emit-llvm -o - | FileCheck %s + +typedef int int4 __attribute__((ext_vector_type(4))); +typedef float float4 __attribute__((ext_vector_type(4))); +typedef _Float16 half; +typedef half half4 __attribute__((ext_vector_type(4))); + +// CHECK-LABEL: define dso_local float @test_builtin_image_load_2d( +// CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0:[0-9]+]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = call float @llvm.amdgcn.image.load.2d.f32.i32.v8i32(i32 12, i32 [[TMP0]], i32 [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], i32 106, i32 103) +// CHECK-NEXT: ret float [[TMP3]] +// +float test_builtin_image_load_2d(float f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_2d_f32_i32(12, i32, i32, tex, 106, 103); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_2d_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP3]] +// +float4 test_builtin_image_load_2d_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_2d_v4f32_i32(100, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_2d_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x half> @llvm.amdgcn.image.load.2d.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP3]] +// +half4 test_builtin_image_load_2d_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_2d_v4f16_i32(100, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_builtin_image_load_2darray( +// CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call float @llvm.amdgcn.image.load.2darray.f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP4]] +// +float test_builtin_image_load_2darray(float f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_2darray_f32_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_2darray_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.load.2darray.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_builtin_image_load_2darray_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_2darray_v4f32_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_2darray_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.load.2darray.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_builtin_image_load_2darray_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_2darray_v4f16_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_1d_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP1]], align 32 +// CHECK-NEXT: [[TMP2:%.*]] = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP2]] +// +float4 test_builtin_image_load_1d_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_1d_v4f32_i32(100, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_1d_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP1]], align 32 +// CHECK-NEXT: [[TMP2:%.*]] = call <4 x half> @llvm.amdgcn.image.load.1d.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP2]] +// +half4 test_builtin_image_load_1d_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_1d_v4f16_i32(100, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_1darray_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.amdgcn.image.load.1darray.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP3]] +// +float4 test_builtin_image_load_1darray_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_1darray_v4f32_i32(100, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_1darray_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x half> @llvm.amdgcn.image.load.1darray.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP3]] +// +half4 test_builtin_image_load_1darray_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_1darray_v4f16_i32(100, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_3d_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_builtin_image_load_3d_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_3d_v4f32_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_3d_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.load.3d.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_builtin_image_load_3d_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_3d_v4f16_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_cube_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.load.cube.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_builtin_image_load_cube_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_cube_v4f32_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_cube_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.load.cube.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_builtin_image_load_cube_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_cube_v4f16_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_mip_1d_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.amdgcn.image.load.mip.1d.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP3]] +// +float4 test_builtin_image_load_mip_1d_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_1d_v4f32_i32(100, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_mip_1d_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x half> @llvm.amdgcn.image.load.mip.1d.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP3]] +// +half4 test_builtin_image_load_mip_1d_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_1d_v4f16_i32(100, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_mip_1darray_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.load.mip.1darray.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_builtin_image_load_mip_1darray_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_1darray_v4f32_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_mip_1darray_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.load.mip.1darray.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_builtin_image_load_mip_1darray_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_1darray_v4f16_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_builtin_image_load_mip_2d( +// CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call float @llvm.amdgcn.image.load.mip.2d.f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP4]] +// +float test_builtin_image_load_mip_2d(float f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_2d_f32_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_mip_2d_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.load.mip.2d.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_builtin_image_load_mip_2d_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_2d_v4f32_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_mip_2d_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.load.mip.2d.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_builtin_image_load_mip_2d_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_2d_v4f16_i32(100, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_builtin_image_load_mip_2darray( +// CHECK-SAME: float noundef [[F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = call float @llvm.amdgcn.image.load.mip.2darray.f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP5]] +// +float test_builtin_image_load_mip_2darray(float f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_2darray_f32_i32(100, i32, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_mip_2darray_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.load.mip.2darray.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_builtin_image_load_mip_2darray_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_2darray_v4f32_i32(100, i32, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_mip_2darray_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.load.mip.2darray.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_builtin_image_load_mip_2darray_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_2darray_v4f16_i32(100, i32, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_mip_3d_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.load.mip.3d.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_builtin_image_load_mip_3d_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_3d_v4f32_i32(100, i32, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_mip_3d_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.load.mip.3d.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_builtin_image_load_mip_3d_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_3d_v4f16_i32(100, i32, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_image_load_mip_cube_1( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.load.mip.cube.v4f32.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_builtin_image_load_mip_cube_1(float4 v4f32, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_cube_v4f32_i32(100, i32, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_image_load_mip_cube_2( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], ptr [[TEX:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP4]], align 32 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.load.mip.cube.v4f16.i32.v8i32(i32 100, i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], <8 x i32> [[TEX_RSRC_VAL]], i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_builtin_image_load_mip_cube_2(half4 v4f16, int i32, __amdgpu_texture_t tex) { + + return __builtin_amdgcn_image_load_mip_cube_v4f16_i32(100, i32, i32, i32, i32, tex, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_amdgcn_image_sample_1d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP1]], align 32 +// CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP2]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP3]] +// +float4 test_builtin_amdgcn_image_sample_1d_v4f32_f32(float4 v4f32, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_1d_v4f32_f32(100, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_amdgcn_image_sample_1d_v4f16_f32( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP1]], align 32 +// CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP3:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.1d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP2]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP3]] +// +half4 test_builtin_amdgcn_image_sample_1d_v4f16_f32(half4 v4f16, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_1d_v4f16_f32(100, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_amdgcn_image_sample_1darray_v4f32_f32( +// CHECK-SAME: i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.1darray.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_builtin_amdgcn_image_sample_1darray_v4f32_f32(int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_1darray_v4f32_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_amdgcn_image_sample_1darray_v4f16_f32( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.1darray.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_builtin_amdgcn_image_sample_1darray_v4f16_f32(half4 v4f16, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_1darray_v4f16_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_builtin_amdgcn_image_sample_2d_f32_f32( +// CHECK-SAME: i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call float @llvm.amdgcn.image.sample.2d.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP4]] +// +float test_builtin_amdgcn_image_sample_2d_f32_f32(int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_2d_f32_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_amdgcn_image_sample_2d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP4]] +// +float4 test_builtin_amdgcn_image_sample_2d_v4f32_f32(float4 v4f32, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_2d_v4f32_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_amdgcn_image_sample_2d_v4f16_f32( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP2]], align 32 +// CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP4:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.2d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP3]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP4]] +// +half4 test_builtin_amdgcn_image_sample_2d_v4f16_f32(half4 v4f16, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_2d_v4f16_f32(100, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local float @test_builtin_amdgcn_image_sample_2darray_f32_f32( +// CHECK-SAME: i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call float @llvm.amdgcn.image.sample.2darray.f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret float [[TMP5]] +// +float test_builtin_amdgcn_image_sample_2darray_f32_f32(int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_2darray_f32_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_amdgcn_image_sample_2darray_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.2darray.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_builtin_amdgcn_image_sample_2darray_v4f32_f32(float4 v4f32, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_2darray_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_amdgcn_image_sample_2darray_v4f16_f32( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.2darray.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_builtin_amdgcn_image_sample_2darray_v4f16_f32(half4 v4f16, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_2darray_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_amdgcn_image_sample_3d_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.3d.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_builtin_amdgcn_image_sample_3d_v4f32_f32(float4 v4f32, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_3d_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_amdgcn_image_sample_3d_v4f16_f32( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.3d.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_builtin_amdgcn_image_sample_3d_v4f16_f32(half4 v4f16, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_3d_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x float> @test_builtin_amdgcn_image_sample_cube_v4f32_f32( +// CHECK-SAME: <4 x float> noundef [[V4F32:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[V4F32_ADDR:%.*]] = alloca <4 x float>, align 16, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F32_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x float> [[V4F32]], ptr [[V4F32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.amdgcn.image.sample.cube.v4f32.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x float> [[TMP5]] +// +float4 test_builtin_amdgcn_image_sample_cube_v4f32_f32(float4 v4f32, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_cube_v4f32_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} + +// CHECK-LABEL: define dso_local <4 x half> @test_builtin_amdgcn_image_sample_cube_v4f16_f32( +// CHECK-SAME: <4 x half> noundef [[V4F16:%.*]], i32 noundef [[I32:%.*]], float noundef [[F32:%.*]], ptr [[TEX:%.*]], <4 x i32> noundef [[VEC4I32:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[RETVAL:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[V4F16_ADDR:%.*]] = alloca <4 x half>, align 8, addrspace(5) +// CHECK-NEXT: [[I32_ADDR:%.*]] = alloca i32, align 4, addrspace(5) +// CHECK-NEXT: [[F32_ADDR:%.*]] = alloca float, align 4, addrspace(5) +// CHECK-NEXT: [[TEX_ADDR:%.*]] = alloca ptr, align 32, addrspace(5) +// CHECK-NEXT: [[VEC4I32_ADDR:%.*]] = alloca <4 x i32>, align 16, addrspace(5) +// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr +// CHECK-NEXT: [[V4F16_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[V4F16_ADDR]] to ptr +// CHECK-NEXT: [[I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I32_ADDR]] to ptr +// CHECK-NEXT: [[F32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F32_ADDR]] to ptr +// CHECK-NEXT: [[TEX_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TEX_ADDR]] to ptr +// CHECK-NEXT: [[VEC4I32_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VEC4I32_ADDR]] to ptr +// CHECK-NEXT: store <4 x half> [[V4F16]], ptr [[V4F16_ADDR_ASCAST]], align 8 +// CHECK-NEXT: store i32 [[I32]], ptr [[I32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store float [[F32]], ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: store ptr [[TEX]], ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: store <4 x i32> [[VEC4I32]], ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[F32_ADDR_ASCAST]], align 4 +// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TEX_ADDR_ASCAST]], align 32 +// CHECK-NEXT: [[TEX_RSRC_VAL:%.*]] = load <8 x i32>, ptr [[TMP3]], align 32 +// CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[VEC4I32_ADDR_ASCAST]], align 16 +// CHECK-NEXT: [[TMP5:%.*]] = call <4 x half> @llvm.amdgcn.image.sample.cube.v4f16.f32.v8i32.v4i32(i32 100, float [[TMP0]], float [[TMP1]], float [[TMP2]], <8 x i32> [[TEX_RSRC_VAL]], <4 x i32> [[TMP4]], i1 false, i32 120, i32 110) +// CHECK-NEXT: ret <4 x half> [[TMP5]] +// +half4 test_builtin_amdgcn_image_sample_cube_v4f16_f32(half4 v4f16, int i32, float f32, __amdgpu_texture_t tex, int4 vec4i32) { + return __builtin_amdgcn_image_sample_cube_v4f16_f32(100, f32, f32, f32, tex, vec4i32, 0, 120, 110); +} |