1 ; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s 2 ; RUN: llc -march=amdgcn -mcpu=tonga -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 6 declare float @llvm.copysign.f32(float, float) nounwind readnone 7 declare <2 x float> @llvm.copysign.v2f32(<2 x float>, <2 x float>) nounwind readnone 8 declare <4 x float> @llvm.copysign.v4f32(<4 x float>, <4 x float>) nounwind readnone 9 10 ; Try to identify arg based on higher address. 11 ; FUNC-LABEL: {{^}}test_copysign_f32: 12 ; SI: s_load_dword [[SMAG:s[0-9]+]], {{.*}} 0xb 13 ; SI: s_load_dword [[SSIGN:s[0-9]+]], {{.*}} 0xc 14 ; VI: s_load_dword [[SMAG:s[0-9]+]], {{.*}} 0x2c 15 ; VI: s_load_dword [[SSIGN:s[0-9]+]], {{.*}} 0x30 16 ; GCN-DAG: v_mov_b32_e32 [[VSIGN:v[0-9]+]], [[SSIGN]] 17 ; GCN-DAG: v_mov_b32_e32 [[VMAG:v[0-9]+]], [[SMAG]] 18 ; GCN-DAG: s_mov_b32 [[SCONST:s[0-9]+]], 0x7fffffff 19 ; GCN: v_bfi_b32 [[RESULT:v[0-9]+]], [[SCONST]], [[VMAG]], [[VSIGN]] 20 ; GCN: buffer_store_dword [[RESULT]], 21 ; GCN: s_endpgm 22 23 ; EG: BFI_INT 24 define void @test_copysign_f32(float addrspace(1)* %out, float %mag, float %sign) nounwind { 25 %result = call float @llvm.copysign.f32(float %mag, float %sign) 26 store float %result, float addrspace(1)* %out, align 4 27 ret void 28 } 29 30 ; FUNC-LABEL: {{^}}test_copysign_v2f32: 31 ; GCN: s_endpgm 32 33 ; EG: BFI_INT 34 ; EG: BFI_INT 35 define void @test_copysign_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %mag, <2 x float> %sign) nounwind { 36 %result = call <2 x float> @llvm.copysign.v2f32(<2 x float> %mag, <2 x float> %sign) 37 store <2 x float> %result, <2 x float> addrspace(1)* %out, align 8 38 ret void 39 } 40 41 ; FUNC-LABEL: {{^}}test_copysign_v4f32: 42 ; GCN: s_endpgm 43 44 ; EG: BFI_INT 45 ; EG: BFI_INT 46 ; EG: BFI_INT 47 ; EG: BFI_INT 48 define void @test_copysign_v4f32(<4 x float> addrspace(1)* %out, <4 x float> %mag, <4 x float> %sign) nounwind { 49 %result = call <4 x float> @llvm.copysign.v4f32(<4 x float> %mag, <4 x float> %sign) 50 store <4 x float> %result, <4 x float> addrspace(1)* %out, align 16 51 ret void 52 } 53 54
