23 lines · plain
1; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji < %s | FileCheck %s2 3declare i32 @llvm.amdgcn.ds.permute(i32, i32) #04 5; CHECK-LABEL: {{^}}ds_permute:6; CHECK: ds_permute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}7define amdgpu_kernel void @ds_permute(ptr addrspace(1) %out, i32 %index, i32 %src) nounwind {8 %bpermute = call i32 @llvm.amdgcn.ds.permute(i32 %index, i32 %src) #09 store i32 %bpermute, ptr addrspace(1) %out, align 410 ret void11}12 13; CHECK-LABEL: {{^}}ds_permute_imm_offset:14; CHECK: ds_permute_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset:415define amdgpu_kernel void @ds_permute_imm_offset(ptr addrspace(1) %out, i32 %base_index, i32 %src) nounwind {16 %index = add i32 %base_index, 417 %bpermute = call i32 @llvm.amdgcn.ds.permute(i32 %index, i32 %src) #018 store i32 %bpermute, ptr addrspace(1) %out, align 419 ret void20}21 22attributes #0 = { nounwind readnone convergent }23