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
333 lines
16 KiB
LLVM
333 lines
16 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -verify-machineinstrs -global-isel -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --check-prefixes=GCN %s
|
|
|
|
declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x4bf16.1k(<4 x i16>, <4 x i16>, <32 x float>, i32, i32, i32)
|
|
declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x4bf16.1k(<4 x i16>, <4 x i16>, <16 x float>, i32, i32, i32)
|
|
declare <4 x float> @llvm.amdgcn.mfma.f32.4x4x4bf16.1k(<4 x i16>, <4 x i16>, <4 x float>, i32, i32, i32)
|
|
declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x8bf16.1k(<4 x i16>, <4 x i16>, <16 x float>, i32, i32, i32)
|
|
declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x16bf16.1k(<4 x i16>, <4 x i16>, <4 x float>, i32, i32, i32)
|
|
declare <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double, double, <4 x double>, i32, i32, i32)
|
|
declare double @llvm.amdgcn.mfma.f64.4x4x4f64(double, double, double, i32, i32, i32)
|
|
declare i32 @llvm.amdgcn.workitem.id.x()
|
|
|
|
define amdgpu_kernel void @test_mfma_f32_32x32x4bf16_1k(ptr addrspace(1) %arg) #0 {
|
|
; GCN-LABEL: test_mfma_f32_32x32x4bf16_1k:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx2 s[34:35], s[4:5], 0x24
|
|
; GCN-NEXT: s_mov_b64 s[36:37], 1
|
|
; GCN-NEXT: v_pk_mov_b32 v[32:33], s[36:37], s[36:37] op_sel:[0,1]
|
|
; GCN-NEXT: s_mov_b32 s36, 2
|
|
; GCN-NEXT: v_pk_mov_b32 v[34:35], s[36:37], s[36:37] op_sel:[0,1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: s_load_dwordx16 s[0:15], s[34:35], 0x0
|
|
; GCN-NEXT: s_load_dwordx16 s[16:31], s[34:35], 0x40
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[8:9], s[8:9], s[8:9] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[10:11], s[10:11], s[10:11] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[12:13], s[12:13], s[12:13] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[14:15], s[14:15], s[14:15] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[16:17], s[16:17], s[16:17] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[20:21], s[20:21], s[20:21] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[22:23], s[22:23], s[22:23] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[24:25], s[24:25], s[24:25] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[26:27], s[26:27], s[26:27] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[28:29], s[28:29], s[28:29] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[30:31], s[30:31], s[30:31] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f32_32x32x4bf16_1k v[0:31], v[32:33], v[34:35], v[0:31] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v32, 0
|
|
; GCN-NEXT: s_nop 15
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[0:3], s[34:35]
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[4:7], s[34:35] offset:16
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[8:11], s[34:35] offset:32
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[12:15], s[34:35] offset:48
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[16:19], s[34:35] offset:64
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[20:23], s[34:35] offset:80
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[24:27], s[34:35] offset:96
|
|
; GCN-NEXT: global_store_dwordx4 v32, v[28:31], s[34:35] offset:112
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%in.1 = load <32 x float>, ptr addrspace(1) %arg
|
|
%a = bitcast i64 1 to <4 x i16>
|
|
%b = bitcast i64 2 to <4 x i16>
|
|
%mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4bf16.1k(<4 x i16> %a, <4 x i16> %b, <32 x float> %in.1, i32 1, i32 2, i32 3)
|
|
store <32 x float> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f32_16x16x4bf16_1k(ptr addrspace(1) %arg) #0 {
|
|
; GCN-LABEL: test_mfma_f32_16x16x4bf16_1k:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx2 s[16:17], s[4:5], 0x24
|
|
; GCN-NEXT: s_mov_b64 s[18:19], 1
|
|
; GCN-NEXT: v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
|
|
; GCN-NEXT: s_mov_b32 s18, 2
|
|
; GCN-NEXT: v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: s_load_dwordx16 s[0:15], s[16:17], 0x0
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[8:9], s[8:9], s[8:9] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[10:11], s[10:11], s[10:11] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[12:13], s[12:13], s[12:13] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[14:15], s[14:15], s[14:15] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f32_16x16x4bf16_1k v[0:15], v[16:17], v[18:19], v[0:15] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v16, 0
|
|
; GCN-NEXT: s_nop 9
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%in.1 = load <16 x float>, ptr addrspace(1) %arg
|
|
%a = bitcast i64 1 to <4 x i16>
|
|
%b = bitcast i64 2 to <4 x i16>
|
|
%mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.16x16x4bf16.1k(<4 x i16> %a, <4 x i16> %b, <16 x float> %in.1, i32 1, i32 2, i32 3)
|
|
store <16 x float> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f32_4x4x4bf16_1k(ptr addrspace(1) %arg) #0 {
|
|
; GCN-LABEL: test_mfma_f32_4x4x4bf16_1k:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24
|
|
; GCN-NEXT: s_mov_b64 s[4:5], 1
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: s_mov_b32 s4, 2
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: s_load_dwordx4 s[0:3], s[6:7], 0x0
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f32_4x4x4bf16_1k v[0:3], v[4:5], v[6:7], v[0:3] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0
|
|
; GCN-NEXT: s_nop 3
|
|
; GCN-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7]
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%in.1 = load <4 x float>, ptr addrspace(1) %arg
|
|
%a = bitcast i64 1 to <4 x i16>
|
|
%b = bitcast i64 2 to <4 x i16>
|
|
%mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x4bf16.1k(<4 x i16> %a, <4 x i16> %b, <4 x float> %in.1, i32 1, i32 2, i32 3)
|
|
store <4 x float> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f32_32x32x8bf16_1k(ptr addrspace(1) %arg) #0 {
|
|
; GCN-LABEL: test_mfma_f32_32x32x8bf16_1k:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx2 s[16:17], s[4:5], 0x24
|
|
; GCN-NEXT: s_mov_b64 s[18:19], 1
|
|
; GCN-NEXT: v_pk_mov_b32 v[16:17], s[18:19], s[18:19] op_sel:[0,1]
|
|
; GCN-NEXT: s_mov_b32 s18, 2
|
|
; GCN-NEXT: v_pk_mov_b32 v[18:19], s[18:19], s[18:19] op_sel:[0,1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: s_load_dwordx16 s[0:15], s[16:17], 0x0
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[8:9], s[8:9], s[8:9] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[10:11], s[10:11], s[10:11] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[12:13], s[12:13], s[12:13] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[14:15], s[14:15], s[14:15] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f32_32x32x8bf16_1k v[0:15], v[16:17], v[18:19], v[0:15] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v16, 0
|
|
; GCN-NEXT: s_nop 15
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
|
|
; GCN-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%in.1 = load <16 x float>, ptr addrspace(1) %arg
|
|
%a = bitcast i64 1 to <4 x i16>
|
|
%b = bitcast i64 2 to <4 x i16>
|
|
%mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x8bf16.1k(<4 x i16> %a, <4 x i16> %b, <16 x float> %in.1, i32 1, i32 2, i32 3)
|
|
store <16 x float> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f32_16x16x16bf16_1k(ptr addrspace(1) %arg) #0 {
|
|
; GCN-LABEL: test_mfma_f32_16x16x16bf16_1k:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24
|
|
; GCN-NEXT: s_mov_b64 s[4:5], 1
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: s_mov_b32 s4, 2
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: s_load_dwordx4 s[0:3], s[6:7], 0x0
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f32_16x16x16bf16_1k v[0:3], v[4:5], v[6:7], v[0:3] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v4, 0
|
|
; GCN-NEXT: s_nop 9
|
|
; GCN-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7]
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%in.1 = load <4 x float>, ptr addrspace(1) %arg
|
|
%a = bitcast i64 1 to <4 x i16>
|
|
%b = bitcast i64 2 to <4 x i16>
|
|
%mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x16bf16.1k(<4 x i16> %a, <4 x i16> %b, <4 x float> %in.1, i32 1, i32 2, i32 3)
|
|
store <4 x float> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f64_4x4x4f64(ptr addrspace(1) %arg, double %a, double %b) #0 {
|
|
; GCN-LABEL: test_mfma_f64_4x4x4f64:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f64_4x4x4f64 v[4:5], v[0:1], v[2:3], 0
|
|
; GCN-NEXT: s_nop 3
|
|
; GCN-NEXT: v_mfma_f64_4x4x4f64 v[0:1], v[0:1], v[2:3], v[4:5] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v2, 0
|
|
; GCN-NEXT: s_nop 7
|
|
; GCN-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1]
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%mai.1 = tail call double @llvm.amdgcn.mfma.f64.4x4x4f64(double %a, double %b, double 0.0, i32 0, i32 0, i32 0)
|
|
%mai.2 = tail call double @llvm.amdgcn.mfma.f64.4x4x4f64(double %a, double %b, double %mai.1, i32 1, i32 2, i32 3)
|
|
store double %mai.2, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f64_16x16x4f64(ptr addrspace(1) %arg, double %a, double %b) #0 {
|
|
; GCN-LABEL: test_mfma_f64_16x16x4f64:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x24
|
|
; GCN-NEXT: s_load_dwordx2 s[12:13], s[4:5], 0x34
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
|
|
; GCN-NEXT: s_load_dwordx8 s[0:7], s[8:9], 0x0
|
|
; GCN-NEXT: v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v8, 0
|
|
; GCN-NEXT: s_nop 15
|
|
; GCN-NEXT: s_nop 0
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[8:9]
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%in.1 = load <4 x double>, ptr addrspace(1) %arg
|
|
%mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %in.1, i32 1, i32 2, i32 3)
|
|
store <4 x double> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm(ptr addrspace(1) %arg, double %a, double %b) #0 {
|
|
; GCN-LABEL: test_mfma_f64_16x16x4f64_splat_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
|
|
; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], 0
|
|
; GCN-NEXT: v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
|
|
; GCN-NEXT: v_mov_b32_e32 v8, 0
|
|
; GCN-NEXT: s_nop 15
|
|
; GCN-NEXT: s_nop 0
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 0.0, double 0.0, double 0.0, double 0.0>, i32 0, i32 0, i32 0)
|
|
%mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
|
|
store <4 x double> %mai.2, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f64_16x16x4f64_imm(ptr addrspace(1) %arg, double %a, double %b) #0 {
|
|
; GCN-LABEL: test_mfma_f64_16x16x4f64_imm:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x24
|
|
; GCN-NEXT: s_load_dwordx2 s[12:13], s[4:5], 0x34
|
|
; GCN-NEXT: s_mov_b64 s[0:1], 0
|
|
; GCN-NEXT: s_mov_b64 s[6:7], 1.0
|
|
; GCN-NEXT: s_mov_b64 s[2:3], s[0:1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
|
|
; GCN-NEXT: s_mov_b64 s[4:5], s[0:1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
|
|
; GCN-NEXT: v_mov_b32_e32 v8, 0
|
|
; GCN-NEXT: s_nop 15
|
|
; GCN-NEXT: s_nop 0
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[8:9]
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 0.0, double 0.0, double 0.0, double 1.0>, i32 0, i32 0, i32 0)
|
|
store <4 x double> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_lit(ptr addrspace(1) %arg, double %a, double %b) #0 {
|
|
; GCN-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
|
|
; GCN: ; %bb.0: ; %bb
|
|
; GCN-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x24
|
|
; GCN-NEXT: s_load_dwordx2 s[12:13], s[4:5], 0x34
|
|
; GCN-NEXT: s_mov_b32 s0, 0
|
|
; GCN-NEXT: s_mov_b32 s1, 0x405ec000
|
|
; GCN-NEXT: s_mov_b64 s[2:3], s[0:1]
|
|
; GCN-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GCN-NEXT: v_pk_mov_b32 v[8:9], s[10:11], s[10:11] op_sel:[0,1]
|
|
; GCN-NEXT: s_mov_b64 s[4:5], s[0:1]
|
|
; GCN-NEXT: s_mov_b64 s[6:7], s[0:1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
|
|
; GCN-NEXT: v_pk_mov_b32 v[10:11], s[12:13], s[12:13] op_sel:[0,1]
|
|
; GCN-NEXT: s_nop 1
|
|
; GCN-NEXT: v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
|
|
; GCN-NEXT: v_mov_b32_e32 v8, 0
|
|
; GCN-NEXT: s_nop 15
|
|
; GCN-NEXT: s_nop 0
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[8:9]
|
|
; GCN-NEXT: global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
|
|
; GCN-NEXT: s_endpgm
|
|
bb:
|
|
%mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 123.0, double 123.0, double 123.0, double 123.0>, i32 0, i32 0, i32 0)
|
|
store <4 x double> %mai.1, ptr addrspace(1) %arg
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { "amdgpu-flat-work-group-size"="1,256" }
|