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
44 lines
3.1 KiB
LLVM
44 lines
3.1 KiB
LLVM
; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -debug-only=machine-scheduler < %s 2>&1 | FileCheck -enable-var-scope %s
|
|
; REQUIRES: asserts
|
|
|
|
declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half>, <4 x half>, <32 x float>, i32, i32, i32)
|
|
|
|
; CHECK: Scheduling SU({{[0-9]+}}) {{.*}} V_MFMA_F32_32X32X4F16
|
|
; CHECK: HWXDL +16x1u
|
|
define amdgpu_kernel void @schedule-xdl-resource(ptr addrspace(1) %in, ptr addrspace(1) %out, ptr addrspace(3) %lds, i32 %stride) #0 {
|
|
%in_ptr.1 = getelementptr <32 x float>, ptr addrspace(1) %in, i32 %stride
|
|
%in_ptr.2 = getelementptr <32 x float>, ptr addrspace(1) %in_ptr.1, i32 %stride
|
|
%in_ptr.3 = getelementptr <32 x float>, ptr addrspace(1) %in_ptr.2, i32 %stride
|
|
%in.load.1 = load <32 x float>, ptr addrspace (1) %in_ptr.1
|
|
%in.load.2 = load <32 x float>, ptr addrspace (1) %in_ptr.2
|
|
%in.load.3 = load <32 x float>, ptr addrspace (1) %in_ptr.3
|
|
%lds_ptr.1 = getelementptr <4 x half>, ptr addrspace(3) %lds, i32 %stride
|
|
%lds_ptr.2 = getelementptr <4 x half>, ptr addrspace(3) %lds_ptr.1, i32 %stride
|
|
%lds_ptr.3 = getelementptr <4 x half>, ptr addrspace(3) %lds_ptr.2, i32 %stride
|
|
%lds.load.1 = load <4 x half>, ptr addrspace(3) %lds_ptr.1
|
|
%lds.load.2 = load <4 x half>, ptr addrspace(3) %lds_ptr.2
|
|
%lds.load.3 = load <4 x half>, ptr addrspace(3) %lds_ptr.3
|
|
%mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half> %lds.load.1, <4 x half> %lds.load.1, <32 x float> %in.load.1, i32 1, i32 1, i32 1)
|
|
%mai.2 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half> %lds.load.2, <4 x half> %lds.load.2, <32 x float> %in.load.2, i32 1, i32 1, i32 1)
|
|
%mai.3 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half> %lds.load.3, <4 x half> %lds.load.3, <32 x float> %in.load.3, i32 1, i32 1, i32 1)
|
|
%mai.4 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half> %lds.load.1, <4 x half> %lds.load.1, <32 x float> %in.load.1, i32 2, i32 2, i32 2)
|
|
%mai.5 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half> %lds.load.2, <4 x half> %lds.load.2, <32 x float> %in.load.2, i32 2, i32 2, i32 2)
|
|
%mai.6 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x4f16(<4 x half> %lds.load.3, <4 x half> %lds.load.3, <32 x float> %in.load.3, i32 2, i32 2, i32 2)
|
|
%out_ptr.1 = getelementptr <32 x float>, ptr addrspace(1) %out, i32 %stride
|
|
%out_ptr.2 = getelementptr <32 x float>, ptr addrspace(1) %out_ptr.1, i32 %stride
|
|
%out_ptr.3 = getelementptr <32 x float>, ptr addrspace(1) %out_ptr.2, i32 %stride
|
|
%out_ptr.4 = getelementptr <32 x float>, ptr addrspace(1) %out_ptr.3, i32 %stride
|
|
%out_ptr.5 = getelementptr <32 x float>, ptr addrspace(1) %out_ptr.4, i32 %stride
|
|
%out_ptr.6 = getelementptr <32 x float>, ptr addrspace(1) %out_ptr.5, i32 %stride
|
|
store <32 x float> %mai.1, ptr addrspace(1) %out_ptr.1
|
|
store <32 x float> %mai.2, ptr addrspace(1) %out_ptr.2
|
|
store <32 x float> %mai.3, ptr addrspace(1) %out_ptr.3
|
|
store <32 x float> %mai.4, ptr addrspace(1) %out_ptr.4
|
|
store <32 x float> %mai.5, ptr addrspace(1) %out_ptr.5
|
|
store <32 x float> %mai.6, ptr addrspace(1) %out_ptr.6
|
|
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind "amdgpu-waves-per-eu"="1,1" }
|