derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
337 lines
16 KiB
LLVM
337 lines
16 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1170 -mattr=+wavefrontsize64 < %s | FileCheck %s --check-prefixes=GCN,GFX1170
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -mattr=+wavefrontsize64 < %s | FileCheck %s --check-prefixes=GCN,GFX12
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm(<4 x half> %A, <4 x half> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_f16_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_f16 v[6:9], v[0:1], v[2:3], 1.0
|
|
; GCN-NEXT: global_store_b128 v[4:5], v[6:9], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_f16_imm_non_inlineable(<4 x half> %A, <4 x half> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_f16_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v6, 0x40400000
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v6
|
|
; GCN-NEXT: v_mov_b32_e32 v8, v6
|
|
; GCN-NEXT: v_mov_b32_e32 v9, v6
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_f16 v[6:9], v[0:1], v[2:3], v[6:9]
|
|
; GCN-NEXT: global_store_b128 v[4:5], v[6:9], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float>@llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half> %A, <4 x half> %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm(<4 x i16> %A, <4 x i16> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_bf16_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_bf16 v[6:9], v[0:1], v[2:3], 1.0
|
|
; GCN-NEXT: global_store_b128 v[4:5], v[6:9], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_bf16_imm_non_inlineable(<4 x i16> %A, <4 x i16> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_bf16_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v6, 0x40400000
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v6
|
|
; GCN-NEXT: v_mov_b32_e32 v8, v6
|
|
; GCN-NEXT: v_mov_b32_e32 v9, v6
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_bf16 v[6:9], v[0:1], v[2:3], v[6:9]
|
|
; GCN-NEXT: global_store_b128 v[4:5], v[6:9], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm(<4 x half> %A, <4 x half> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f16_16x16x16_f16_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_f16_16x16x16_f16 v[6:7], v[0:1], v[2:3], 1.0
|
|
; GCN-NEXT: global_store_b64 v[4:5], v[6:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> <half 1.0, half 1.0, half 1.0, half 1.0>, i1 0)
|
|
store <4 x half> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f16_16x16x16_f16_imm_non_inlineable(<4 x half> %A, <4 x half> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f16_16x16x16_f16_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v6, 0x42004200
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v6
|
|
; GCN-NEXT: v_wmma_f16_16x16x16_f16 v[6:7], v[0:1], v[2:3], v[6:7]
|
|
; GCN-NEXT: global_store_b64 v[4:5], v[6:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half> %A, <4 x half> %B, <4 x half> <half 3.0, half 3.0, half 3.0, half 3.0>, i1 0)
|
|
store <4 x half> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm(<4 x i16> %A, <4 x i16> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_bf16_16x16x16_bf16_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v6, 0x3f803f80
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v6
|
|
; GCN-NEXT: v_wmma_bf16_16x16x16_bf16 v[6:7], v[0:1], v[2:3], v[6:7]
|
|
; GCN-NEXT: global_store_b64 v[4:5], v[6:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> <i16 16256, i16 16256, i16 16256, i16 16256>, i1 0)
|
|
store <4 x i16> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_imm_non_inlineable(<4 x i16> %A, <4 x i16> %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_bf16_16x16x16_bf16_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v6, 0x3fc03fc0
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v6
|
|
; GCN-NEXT: v_wmma_bf16_16x16x16_bf16 v[6:7], v[0:1], v[2:3], v[6:7]
|
|
; GCN-NEXT: global_store_b64 v[4:5], v[6:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16> %A, <4 x i16> %B, <4 x i16> <i16 16320, i16 16320, i16 16320, i16 16320>, i1 0)
|
|
store <4 x i16> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm(i32 %A, i32 %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_i32_16x16x16_iu8_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_i32_16x16x16_iu8 v[4:7], v0, v1, 1
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
|
|
store <4 x i32> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_i32_16x16x16_iu8_imm_non_inlineable(i32 %A, i32 %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_i32_16x16x16_iu8_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0x80
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v5, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v6, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v4
|
|
; GCN-NEXT: v_wmma_i32_16x16x16_iu8 v[4:7], v0, v1, v[4:7]
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
|
|
store <4 x i32> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm(i32 %A, i32 %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_i32_16x16x16_iu4_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_i32_16x16x16_iu4 v[4:7], v0, v1, 1
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
|
|
store <4 x i32> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_i32_16x16x16_iu4_imm_non_inlineable(i32 %A, i32 %B, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_i32_16x16x16_iu4_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0x80
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v5, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v6, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v4
|
|
; GCN-NEXT: v_wmma_i32_16x16x16_iu4 v[4:7], v0, v1, v[4:7]
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
|
|
store <4 x i32> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_fp8_fp8_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_fp8_fp8 v[4:7], v0, v1, 1.0
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_fp8_fp8_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0x40400000
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v5, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v6, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v4
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_fp8_fp8 v[4:7], v0, v1, v[4:7]
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_bf8_fp8_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_bf8_fp8 v[4:7], v0, v1, 1.0
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_bf8_fp8_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0x40400000
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v5, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v6, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v4
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_bf8_fp8 v[4:7], v0, v1, v[4:7]
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_fp8_bf8_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_fp8_bf8 v[4:7], v0, v1, 1.0
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_fp8_bf8_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0x40400000
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v5, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v6, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v4
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_fp8_bf8 v[4:7], v0, v1, v[4:7]
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_bf8_bf8_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_bf8_bf8 v[4:7], v0, v1, 1.0
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable(i32 %A, i32 %B, <4 x float> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_f32_16x16x16_bf8_bf8_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0x40400000
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v5, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v6, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v4
|
|
; GCN-NEXT: v_wmma_f32_16x16x16_bf8_bf8 v[4:7], v0, v1, v[4:7]
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32 %A, i32 %B, <4 x float> <float 3.0, float 3.0, float 3.0, float 3.0>)
|
|
store <4 x float> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm(i32 %A, i32 %B, <4 x i32> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_i32_16x16x32_iu4_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_wmma_i32_16x16x32_iu4 v[4:7], v0, v1, 1
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, i1 0)
|
|
store <4 x i32> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @test_wmma_i32_16x16x32_iu4_imm_non_inlineable(i32 %A, i32 %B, <4 x i32> %C, ptr addrspace(1) %out) {
|
|
; GCN-LABEL: test_wmma_i32_16x16x32_iu4_imm_non_inlineable:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0x80
|
|
; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GCN-NEXT: v_mov_b32_e32 v5, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v6, v4
|
|
; GCN-NEXT: v_mov_b32_e32 v7, v4
|
|
; GCN-NEXT: v_wmma_i32_16x16x32_iu4 v[4:7], v0, v1, v[4:7]
|
|
; GCN-NEXT: global_store_b128 v[2:3], v[4:7], off
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 0, i32 %A, i1 0, i32 %B, <4 x i32> <i32 128, i32 128, i32 128, i32 128>, i1 0)
|
|
store <4 x i32> %res, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v4f32.v4f16(<4 x half>, <4 x half>, <4 x float>)
|
|
declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v4f32.v4i16(<4 x i16>, <4 x i16>, <4 x float>)
|
|
declare <4 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v4f16.v4f16(<4 x half>, <4 x half>, <4 x half>, i1 immarg)
|
|
declare <4 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v4i16.v4i16(<4 x i16>, <4 x i16>, <4 x i16>, i1 immarg)
|
|
declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
|
|
declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
|
|
declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.fp8.v4f32.i32(i32, i32, <4 x float>)
|
|
declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.fp8.bf8.v4f32.i32(i32, i32, <4 x float>)
|
|
declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.fp8.v4f32.i32(i32, i32, <4 x float>)
|
|
declare <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf8.bf8.v4f32.i32(i32, i32, <4 x float>)
|
|
declare <4 x i32> @llvm.amdgcn.wmma.i32.16x16x32.iu4.v4i32.i32(i1 immarg, i32, i1 immarg, i32, <4 x i32>, i1 immarg)
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; GFX1170: {{.*}}
|
|
; GFX12: {{.*}}
|