1 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s 2 ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s 3 ; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s 4 5 declare float @llvm.copysign.f32(float, float) nounwind readnone 6 declare <2 x float> @llvm.copysign.v2f32(<2 x float>, <2 x float>) nounwind readnone 7 declare <4 x float> @llvm.copysign.v4f32(<4 x float>, <4 x float>) nounwind readnone 8 9 ; Try to identify arg based on higher address. 10 ; FUNC-LABEL: {{^}}test_copysign_f32: 11 ; SI: s_load_dwordx2 s{{\[}}[[SMAG:[0-9]+]]:[[SSIGN:[0-9]+]]{{\]}}, {{.*}} 0xb 12 ; VI: s_load_dwordx2 s{{\[}}[[SMAG:[0-9]+]]:[[SSIGN:[0-9]+]]{{\]}}, {{.*}} 0x2c 13 14 ; GCN-DAG: v_mov_b32_e32 [[VSIGN:v[0-9]+]], s[[SSIGN]] 15 ; GCN-DAG: v_mov_b32_e32 [[VMAG:v[0-9]+]], s[[SMAG]] 16 ; GCN-DAG: s_brev_b32 [[SCONST:s[0-9]+]], -2 17 ; GCN: v_bfi_b32 [[RESULT:v[0-9]+]], [[SCONST]], [[VMAG]], [[VSIGN]] 18 ; GCN: buffer_store_dword [[RESULT]], 19 ; GCN: s_endpgm 20 21 ; EG: BFI_INT 22 define amdgpu_kernel void @test_copysign_f32(float addrspace(1)* %out, float %mag, float %sign) nounwind { 23 %result = call float @llvm.copysign.f32(float %mag, float %sign) 24 store float %result, float addrspace(1)* %out, align 4 25 ret void 26 } 27 28 ; FUNC-LABEL: {{^}}test_copysign_v2f32: 29 ; GCN: s_endpgm 30 31 ; EG: BFI_INT 32 ; EG: BFI_INT 33 define amdgpu_kernel void @test_copysign_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %mag, <2 x float> %sign) nounwind { 34 %result = call <2 x float> @llvm.copysign.v2f32(<2 x float> %mag, <2 x float> %sign) 35 store <2 x float> %result, <2 x float> addrspace(1)* %out, align 8 36 ret void 37 } 38 39 ; FUNC-LABEL: {{^}}test_copysign_v4f32: 40 ; GCN: s_endpgm 41 42 ; EG: BFI_INT 43 ; EG: BFI_INT 44 ; EG: BFI_INT 45 ; EG: BFI_INT 46 define amdgpu_kernel void @test_copysign_v4f32(<4 x float> addrspace(1)* %out, <4 x float> %mag, <4 x float> %sign) nounwind { 47 %result = call <4 x float> @llvm.copysign.v4f32(<4 x float> %mag, <4 x float> %sign) 48 store <4 x float> %result, <4 x float> addrspace(1)* %out, align 16 49 ret void 50 } 51 52
