33 lines · plain
1; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 --amdgpu-memcpy-loop-unroll=100000 < %s | FileCheck --check-prefixes=GCN,GFX11 %s2; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1200 --amdgpu-memcpy-loop-unroll=100000 < %s | FileCheck --check-prefixes=GCN,GFX12 %s3 4; GCN-LABEL: .amdhsa_kernel large5; GFX11: .amdhsa_inst_pref_size 36; GFX11: codeLenInByte = 3{{[0-9][0-9]$}}7; GFX12: .amdhsa_inst_pref_size 48; GFX12: codeLenInByte = 4{{[0-9][0-9]$}}9define amdgpu_kernel void @large(ptr addrspace(1) %out, ptr addrspace(1) %in) {10bb:11 call void @llvm.memcpy.p1.p3.i32(ptr addrspace(1) %out, ptr addrspace(1) %in, i32 256, i1 false)12 ret void13}14 15; GCN-LABEL: .amdhsa_kernel small16; GCN: .amdhsa_inst_pref_size 117; GCN: codeLenInByte = {{[0-9]$}}18define amdgpu_kernel void @small() {19bb:20 ret void21}22 23; Ignore inline asm in size calculation24 25; GCN-LABEL: .amdhsa_kernel inline_asm26; GCN: .amdhsa_inst_pref_size 127; GCN: codeLenInByte = {{[0-9]$}}28define amdgpu_kernel void @inline_asm() {29bb:30 call void asm sideeffect ".fill 256, 4, 0", ""()31 ret void32}33