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
4 declare double @llvm.copysign.f64(double, double) nounwind readnone
5 declare <2 x double> @llvm.copysign.v2f64(<2 x double>, <2 x double>) nounwind readnone
6 declare <4 x double> @llvm.copysign.v4f64(<4 x double>, <4 x double>) nounwind readnone
8 ; FUNC-LABEL: {{^}}test_copysign_f64:
9 ; SI-DAG: s_load_dwordx2 s{{\[}}[[SMAG_LO:[0-9]+]]:[[SMAG_HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x13
10 ; SI-DAG: s_load_dwordx2 s{{\[}}[[SSIGN_LO:[0-9]+]]:[[SSIGN_HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x1d
11 ; VI-DAG: s_load_dwordx2 s{{\[}}[[SMAG_LO:[0-9]+]]:[[SMAG_HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x4c
12 ; VI-DAG: s_load_dwordx2 s{{\[}}[[SSIGN_LO:[0-9]+]]:[[SSIGN_HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x74
13 ; GCN-DAG: v_mov_b32_e32 v[[VSIGN_HI:[0-9]+]], s[[SSIGN_HI]]
14 ; GCN-DAG: v_mov_b32_e32 v[[VMAG_HI:[0-9]+]], s[[SMAG_HI]]
15 ; GCN-DAG: s_brev_b32 [[SCONST:s[0-9]+]], -2
16 ; GCN-DAG: v_bfi_b32 v[[VRESULT_HI:[0-9]+]], [[SCONST]], v[[VMAG_HI]], v[[VSIGN_HI]]
17 ; GCN-DAG: v_mov_b32_e32 v[[VMAG_LO:[0-9]+]], s[[SMAG_LO]]
18 ; GCN: buffer_store_dwordx2 v{{\[}}[[VMAG_LO]]:[[VRESULT_HI]]{{\]}}
20 define amdgpu_kernel void @test_copysign_f64(double addrspace(1)* %out, [8 x i32], double %mag, [8 x i32], double %sign) nounwind {
21 %result = call double @llvm.copysign.f64(double %mag, double %sign)
22 store double %result, double addrspace(1)* %out, align 8
26 ; FUNC-LABEL: {{^}}test_copysign_f64_f32:
27 ; SI-DAG: s_load_dwordx2 s{{\[}}[[SMAG_LO:[0-9]+]]:[[SMAG_HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x13
28 ; VI-DAG: s_load_dwordx2 s{{\[}}[[SMAG_LO:[0-9]+]]:[[SMAG_HI:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x4c
29 ; GCN-DAG: s_load_dword s[[SSIGN:[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}
30 ; GCN-DAG: s_brev_b32 [[SCONST:s[0-9]+]], -2{{$}}
31 ; GCN-DAG: v_mov_b32_e32 v[[VMAG_HI:[0-9]+]], s[[SMAG_HI]]
32 ; GCN-DAG: v_mov_b32_e32 v[[VSIGN:[0-9]+]], s[[SSIGN]]
33 ; GCN-DAG: v_bfi_b32 v[[VRESULT_HI:[0-9]+]], [[SCONST]], v[[VMAG_HI]], v[[VSIGN]]
34 ; GCN-DAG: v_mov_b32_e32 v[[VMAG_LO:[0-9]+]], s[[SMAG_LO]]
35 ; GCN: buffer_store_dwordx2 v{{\[}}[[VMAG_LO]]:[[VRESULT_HI]]{{\]}}
36 define amdgpu_kernel void @test_copysign_f64_f32(double addrspace(1)* %out, [8 x i32], double %mag, float %sign) nounwind {
37 %c = fpext float %sign to double
38 %result = call double @llvm.copysign.f64(double %mag, double %c)
39 store double %result, double addrspace(1)* %out, align 8
43 ; FUNC-LABEL: {{^}}test_copysign_v2f64:
45 define amdgpu_kernel void @test_copysign_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %mag, <2 x double> %sign) nounwind {
46 %result = call <2 x double> @llvm.copysign.v2f64(<2 x double> %mag, <2 x double> %sign)
47 store <2 x double> %result, <2 x double> addrspace(1)* %out, align 8
51 ; FUNC-LABEL: {{^}}test_copysign_v4f64:
53 define amdgpu_kernel void @test_copysign_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %mag, <4 x double> %sign) nounwind {
54 %result = call <4 x double> @llvm.copysign.v4f64(<4 x double> %mag, <4 x double> %sign)
55 store <4 x double> %result, <4 x double> addrspace(1)* %out, align 8