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
73 lines
3.5 KiB
LLVM
73 lines
3.5 KiB
LLVM
; RUN: llc -mtriple=amdgcn -mattr=+load-store-opt < %s | FileCheck -check-prefix=SI --check-prefix=CHECK %s
|
|
; RUN: llc -mtriple=amdgcn -mcpu=bonaire -mattr=+load-store-opt < %s | FileCheck -check-prefix=CI --check-prefix=CHECK %s
|
|
; RUN: llc -mtriple=amdgcn -mattr=+load-store-opt,+unsafe-ds-offset-folding < %s | FileCheck -check-prefix=CI --check-prefix=CHECK %s
|
|
|
|
declare i32 @llvm.amdgcn.workitem.id.x() #0
|
|
declare void @llvm.amdgcn.s.barrier() #1
|
|
|
|
; Function Attrs: nounwind
|
|
; CHECK-LABEL: {{^}}signed_ds_offset_addressing_loop:
|
|
; CHECK: BB0_1:
|
|
; CHECK: v_add_i32_e32 [[VADDR:v[0-9]+]],
|
|
; SI-DAG: v_add_i32_e32 [[VADDR1:v[0-9]+]], vcc, 0xc20, [[VADDR]]
|
|
; SI-DAG: ds_read_b32 v{{[0-9]+}}, [[VADDR1]]
|
|
; SI-DAG: v_add_i32_e32 [[VADDR2:v[0-9]+]], vcc, {{s[0-9]+}}, [[VADDR]]
|
|
; SI-DAG: ds_read_b32 v{{[0-9]+}}, [[VADDR2]]
|
|
; SI-DAG: v_add_i32_e32 [[VADDR3:v[0-9]+]], vcc, 0xca0, [[VADDR]]
|
|
; SI-DAG: ds_read_b32 v{{[0-9]+}}, [[VADDR3]]
|
|
; SI-DAG: v_add_i32_e32 [[VADDR4:v[0-9]+]], vcc, 0xca8, [[VADDR]]
|
|
; SI-DAG: ds_read_b32 v{{[0-9]+}}, [[VADDR4]]
|
|
; SI-DAG: v_add_i32_e32 [[VADDR5:v[0-9]+]], vcc, 0xd20, [[VADDR]]
|
|
; SI-DAG: ds_read_b32 v{{[0-9]+}}, [[VADDR5]]
|
|
;
|
|
; CI: v_add_i32_e32 [[VADDRCI:v[0-9]+]], vcc, {{s[0-9]+}}, [[VADDR]]
|
|
; CI-DAG: ds_read2_b32 v{{\[[0-9]+:[0-9]+\]}}, [[VADDRCI]] offset0:8 offset1:10
|
|
; CI-DAG: ds_read2_b32 v{{\[[0-9]+:[0-9]+\]}}, [[VADDRCI]] offset0:40 offset1:42
|
|
; CI-DAG: ds_read_b32 v{{[0-9]+}}, [[VADDR]] offset:3360
|
|
; CHECK: s_endpgm
|
|
define amdgpu_kernel void @signed_ds_offset_addressing_loop(ptr addrspace(1) noalias nocapture %out, ptr addrspace(3) noalias nocapture readonly %lptr, i32 %n) #2 {
|
|
entry:
|
|
%x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #0
|
|
%mul = shl nsw i32 %x.i, 1
|
|
br label %for.body
|
|
|
|
for.body: ; preds = %for.body, %entry
|
|
%sum.03 = phi float [ 0.000000e+00, %entry ], [ %add13, %for.body ]
|
|
%offset.02 = phi i32 [ %mul, %entry ], [ %add14, %for.body ]
|
|
%k.01 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
tail call void @llvm.amdgcn.s.barrier() #1
|
|
%arrayidx = getelementptr inbounds float, ptr addrspace(3) %lptr, i32 %offset.02
|
|
%tmp = load float, ptr addrspace(3) %arrayidx, align 4
|
|
%add1 = add nsw i32 %offset.02, 2
|
|
%arrayidx2 = getelementptr inbounds float, ptr addrspace(3) %lptr, i32 %add1
|
|
%tmp1 = load float, ptr addrspace(3) %arrayidx2, align 4
|
|
%add3 = add nsw i32 %offset.02, 32
|
|
%arrayidx4 = getelementptr inbounds float, ptr addrspace(3) %lptr, i32 %add3
|
|
%tmp2 = load float, ptr addrspace(3) %arrayidx4, align 4
|
|
%add5 = add nsw i32 %offset.02, 34
|
|
%arrayidx6 = getelementptr inbounds float, ptr addrspace(3) %lptr, i32 %add5
|
|
%tmp3 = load float, ptr addrspace(3) %arrayidx6, align 4
|
|
%add7 = add nsw i32 %offset.02, 64
|
|
%arrayidx8 = getelementptr inbounds float, ptr addrspace(3) %lptr, i32 %add7
|
|
%tmp4 = load float, ptr addrspace(3) %arrayidx8, align 4
|
|
%add9 = fadd float %tmp, %tmp1
|
|
%add10 = fadd float %add9, %tmp2
|
|
%add11 = fadd float %add10, %tmp3
|
|
%add12 = fadd float %add11, %tmp4
|
|
%add13 = fadd float %sum.03, %add12
|
|
%inc = add nsw i32 %k.01, 1
|
|
%add14 = add nsw i32 %offset.02, 97
|
|
%exitcond = icmp eq i32 %inc, 8
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
for.end: ; preds = %for.body
|
|
%tmp5 = sext i32 %x.i to i64
|
|
%arrayidx15 = getelementptr inbounds float, ptr addrspace(1) %out, i64 %tmp5
|
|
store float %add13, ptr addrspace(1) %arrayidx15, align 4
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind readnone }
|
|
attributes #1 = { convergent nounwind }
|
|
attributes #2 = { nounwind }
|