Skip to content

[AMDGPU] Add intrinsics for v_[pk]_add_{min|max}_* instructions #164731

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 23, 2025
Merged

[AMDGPU] Add intrinsics for v_[pk]_add_{min|max}_* instructions #164731

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions clang/include/clang/Basic/BuiltinsAMDGPU.def
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -830,6 +830,15 @@ TARGET_BUILTIN(__builtin_amdgcn_perm_pk16_b4_u4, "V2UiUiUiV2Ui", "nc", "tensor-c
TARGET_BUILTIN(__builtin_amdgcn_perm_pk16_b6_u4, "V3UiUiULiV2Ui", "nc", "tensor-cvt-lut-insts")
TARGET_BUILTIN(__builtin_amdgcn_perm_pk16_b8_u4, "V4UiULiULiV2Ui", "nc", "tensor-cvt-lut-insts")

TARGET_BUILTIN(__builtin_amdgcn_add_max_i32, "iiiiIb", "nc", "add-min-max-insts")
TARGET_BUILTIN(__builtin_amdgcn_add_max_u32, "UiUiUiUiIb", "nc", "add-min-max-insts")
TARGET_BUILTIN(__builtin_amdgcn_add_min_i32, "iiiiIb", "nc", "add-min-max-insts")
TARGET_BUILTIN(__builtin_amdgcn_add_min_u32, "UiUiUiUiIb", "nc", "add-min-max-insts")
TARGET_BUILTIN(__builtin_amdgcn_pk_add_max_i16, "V2sV2sV2sV2sIb", "nc", "pk-add-min-max-insts")
TARGET_BUILTIN(__builtin_amdgcn_pk_add_max_u16, "V2UsV2UsV2UsV2UsIb", "nc", "pk-add-min-max-insts")
TARGET_BUILTIN(__builtin_amdgcn_pk_add_min_i16, "V2sV2sV2sV2sIb", "nc", "pk-add-min-max-insts")
TARGET_BUILTIN(__builtin_amdgcn_pk_add_min_u16, "V2UsV2UsV2UsV2UsIb", "nc", "pk-add-min-max-insts")

// GFX1250 WMMA builtins
TARGET_BUILTIN(__builtin_amdgcn_wmma_f32_16x16x4_f32, "V8fIbV2fIbV2fIsV8fIbIb", "nc", "gfx1250-insts,wavefrontsize32")
TARGET_BUILTIN(__builtin_amdgcn_wmma_f32_16x16x32_bf16, "V8fIbV16yIbV16yIsV8fIbIb", "nc", "gfx1250-insts,wavefrontsize32")
Expand Down
4 changes: 2 additions & 2 deletions clang/test/CodeGenOpenCL/amdgpu-cluster-dims.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,8 +26,8 @@ kernel void foo(global int *p) { *p = 1; }
// CHECK-NEXT: ret void
//
//.
// CHECK: attributes #[[ATTR0]] = { convergent norecurse nounwind "amdgpu-cluster-dims"="0,0,0" "amdgpu-flat-work-group-size"="1,256" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx1250" "target-features"="+16-bit-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32" "uniform-work-group-size"="false" }
// CHECK: attributes #[[ATTR1]] = { alwaysinline convergent norecurse nounwind "amdgpu-cluster-dims"="0,0,0" "amdgpu-flat-work-group-size"="1,256" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx1250" "target-features"="+16-bit-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32" }
// CHECK: attributes #[[ATTR0]] = { convergent norecurse nounwind "amdgpu-cluster-dims"="0,0,0" "amdgpu-flat-work-group-size"="1,256" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx1250" "target-features"="+16-bit-insts,+add-min-max-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+pk-add-min-max-insts,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32" "uniform-work-group-size"="false" }
// CHECK: attributes #[[ATTR1]] = { alwaysinline convergent norecurse nounwind "amdgpu-cluster-dims"="0,0,0" "amdgpu-flat-work-group-size"="1,256" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx1250" "target-features"="+16-bit-insts,+add-min-max-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+pk-add-min-max-insts,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32" }
// CHECK: attributes #[[ATTR2]] = { convergent nounwind }
//.
// CHECK: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 600}
Expand Down
4 changes: 2 additions & 2 deletions clang/test/CodeGenOpenCL/amdgpu-features.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -109,8 +109,8 @@
// GFX1153: "target-features"="+16-bit-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f32,+ci-insts,+dl-insts,+dot10-insts,+dot12-insts,+dot5-insts,+dot7-insts,+dot8-insts,+dot9-insts,+dpp,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize32"
// GFX1200: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot10-insts,+dot11-insts,+dot12-insts,+dot7-insts,+dot8-insts,+dot9-insts,+dpp,+fp8-conversion-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize32"
// GFX1201: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot10-insts,+dot11-insts,+dot12-insts,+dot7-insts,+dot8-insts,+dot9-insts,+dpp,+fp8-conversion-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize32"
// GFX1250: "target-features"="+16-bit-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32"
// GFX1251: "target-features"="+16-bit-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32"
// GFX1250: "target-features"="+16-bit-insts,+add-min-max-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+pk-add-min-max-insts,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32"
// GFX1251: "target-features"="+16-bit-insts,+add-min-max-insts,+ashr-pk-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-buffer-pk-add-bf16-inst,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-fmin-fmax-global-f32,+atomic-fmin-fmax-global-f64,+atomic-global-pk-add-bf16-inst,+bf16-cvt-insts,+bf16-pk-insts,+bf16-trans-insts,+bitop3-insts,+ci-insts,+clusters,+dl-insts,+dot7-insts,+dot8-insts,+dpp,+fp8-conversion-insts,+fp8e5m3-insts,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx12-insts,+gfx1250-insts,+gfx8-insts,+gfx9-insts,+permlane16-swap,+pk-add-min-max-insts,+prng-inst,+setprio-inc-wg-inst,+tanh-insts,+tensor-cvt-lut-insts,+transpose-load-f4f6-insts,+vmem-pref-insts,+wavefrontsize32"

// GFX1103-W64: "target-features"="+16-bit-insts,+atomic-fadd-rtn-insts,+atomic-fmin-fmax-global-f32,+ci-insts,+dl-insts,+dot10-insts,+dot12-insts,+dot5-insts,+dot7-insts,+dot8-insts,+dot9-insts,+dpp,+gfx10-3-insts,+gfx10-insts,+gfx11-insts,+gfx8-insts,+gfx9-insts,+wavefrontsize64"

Expand Down
109 changes: 109 additions & 0 deletions clang/test/CodeGenOpenCL/builtins-amdgcn-gfx1250.cl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ typedef float __attribute__((ext_vector_type(8))) float8;
typedef float __attribute__((ext_vector_type(16))) float16;
typedef float __attribute__((ext_vector_type(32))) float32;
typedef short __attribute__((ext_vector_type(2))) short2;
typedef unsigned short __attribute__((ext_vector_type(2))) ushort2;

// CHECK-LABEL: @test_setprio_inc_wg(
// CHECK-NEXT: entry:
Expand Down Expand Up @@ -1718,3 +1719,111 @@ void test_cvt_f32_fp8_e5m3(global int* out, int a)
*out = __builtin_amdgcn_cvt_f32_fp8_e5m3(a, 2);
*out = __builtin_amdgcn_cvt_f32_fp8_e5m3(a, 3);
}

// CHECK-LABEL: @test_add_min_max(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[C_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[OUT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[OUT_ADDR]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
// CHECK-NEXT: [[C_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[C_ADDR]] to ptr
// CHECK-NEXT: store ptr addrspace(1) [[OUT:%.*]], ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: store i32 [[B:%.*]], ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: store i32 [[C:%.*]], ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @llvm.amdgcn.add.max.i32(i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i1 false)
// CHECK-NEXT: [[TMP4:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[TMP3]], ptr addrspace(1) [[TMP4]], align 4
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP8:%.*]] = call i32 @llvm.amdgcn.add.max.u32(i32 [[TMP5]], i32 [[TMP6]], i32 [[TMP7]], i1 true)
// CHECK-NEXT: [[TMP9:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[TMP8]], ptr addrspace(1) [[TMP9]], align 4
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @llvm.amdgcn.add.min.i32(i32 [[TMP10]], i32 [[TMP11]], i32 [[TMP12]], i1 false)
// CHECK-NEXT: [[TMP14:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[TMP13]], ptr addrspace(1) [[TMP14]], align 4
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP18:%.*]] = call i32 @llvm.amdgcn.add.min.u32(i32 [[TMP15]], i32 [[TMP16]], i32 [[TMP17]], i1 true)
// CHECK-NEXT: [[TMP19:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[TMP18]], ptr addrspace(1) [[TMP19]], align 4
// CHECK-NEXT: ret void
//
void test_add_min_max(global int *out, int a, int b, int c)
{
*out = __builtin_amdgcn_add_max_i32(a, b, c, false);
*out = __builtin_amdgcn_add_max_u32(a, b, c, true);
*out = __builtin_amdgcn_add_min_i32(a, b, c, false);
*out = __builtin_amdgcn_add_min_u32(a, b, c, true);
}

// CHECK-LABEL: @test_pk_add_min_max(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// CHECK-NEXT: [[UOUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca <2 x i16>, align 4, addrspace(5)
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca <2 x i16>, align 4, addrspace(5)
// CHECK-NEXT: [[C_ADDR:%.*]] = alloca <2 x i16>, align 4, addrspace(5)
// CHECK-NEXT: [[UA_ADDR:%.*]] = alloca <2 x i16>, align 4, addrspace(5)
// CHECK-NEXT: [[UB_ADDR:%.*]] = alloca <2 x i16>, align 4, addrspace(5)
// CHECK-NEXT: [[UC_ADDR:%.*]] = alloca <2 x i16>, align 4, addrspace(5)
// CHECK-NEXT: [[OUT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[OUT_ADDR]] to ptr
// CHECK-NEXT: [[UOUT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[UOUT_ADDR]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
// CHECK-NEXT: [[C_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[C_ADDR]] to ptr
// CHECK-NEXT: [[UA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[UA_ADDR]] to ptr
// CHECK-NEXT: [[UB_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[UB_ADDR]] to ptr
// CHECK-NEXT: [[UC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[UC_ADDR]] to ptr
// CHECK-NEXT: store ptr addrspace(1) [[OUT:%.*]], ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store ptr addrspace(1) [[UOUT:%.*]], ptr [[UOUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store <2 x i16> [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: store <2 x i16> [[B:%.*]], ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: store <2 x i16> [[C:%.*]], ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: store <2 x i16> [[UA:%.*]], ptr [[UA_ADDR_ASCAST]], align 4
// CHECK-NEXT: store <2 x i16> [[UB:%.*]], ptr [[UB_ADDR_ASCAST]], align 4
// CHECK-NEXT: store <2 x i16> [[UC:%.*]], ptr [[UC_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load <2 x i16>, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = load <2 x i16>, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = load <2 x i16>, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = call <2 x i16> @llvm.amdgcn.pk.add.max.i16(<2 x i16> [[TMP0]], <2 x i16> [[TMP1]], <2 x i16> [[TMP2]], i1 false)
// CHECK-NEXT: [[TMP4:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store <2 x i16> [[TMP3]], ptr addrspace(1) [[TMP4]], align 4
// CHECK-NEXT: [[TMP5:%.*]] = load <2 x i16>, ptr [[UA_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP6:%.*]] = load <2 x i16>, ptr [[UB_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP7:%.*]] = load <2 x i16>, ptr [[UC_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP8:%.*]] = call <2 x i16> @llvm.amdgcn.pk.add.max.u16(<2 x i16> [[TMP5]], <2 x i16> [[TMP6]], <2 x i16> [[TMP7]], i1 true)
// CHECK-NEXT: [[TMP9:%.*]] = load ptr addrspace(1), ptr [[UOUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store <2 x i16> [[TMP8]], ptr addrspace(1) [[TMP9]], align 4
// CHECK-NEXT: [[TMP10:%.*]] = load <2 x i16>, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP11:%.*]] = load <2 x i16>, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP12:%.*]] = load <2 x i16>, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP13:%.*]] = call <2 x i16> @llvm.amdgcn.pk.add.min.i16(<2 x i16> [[TMP10]], <2 x i16> [[TMP11]], <2 x i16> [[TMP12]], i1 false)
// CHECK-NEXT: [[TMP14:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store <2 x i16> [[TMP13]], ptr addrspace(1) [[TMP14]], align 4
// CHECK-NEXT: [[TMP15:%.*]] = load <2 x i16>, ptr [[UA_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP16:%.*]] = load <2 x i16>, ptr [[UB_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP17:%.*]] = load <2 x i16>, ptr [[UC_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP18:%.*]] = call <2 x i16> @llvm.amdgcn.pk.add.min.u16(<2 x i16> [[TMP15]], <2 x i16> [[TMP16]], <2 x i16> [[TMP17]], i1 true)
// CHECK-NEXT: [[TMP19:%.*]] = load ptr addrspace(1), ptr [[UOUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store <2 x i16> [[TMP18]], ptr addrspace(1) [[TMP19]], align 4
// CHECK-NEXT: ret void
//
void test_pk_add_min_max(global short2 *out, global ushort2 *uout, short2 a, short2 b, short2 c, ushort2 ua, ushort2 ub, ushort2 uc)
{
*out = __builtin_amdgcn_pk_add_max_i16(a, b, c, false);
*uout = __builtin_amdgcn_pk_add_max_u16(ua, ub, uc, true);
*out = __builtin_amdgcn_pk_add_min_i16(a, b, c, false);
*uout = __builtin_amdgcn_pk_add_min_u16(ua, ub, uc, true);
}
Loading