1 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s 2 3 ; GCN-LABEL: {{^}}test_default_si: 4 ; GCN: FloatMode: 192 5 ; GCN: IeeeMode: 1 6 define amdgpu_kernel void @test_default_si(float addrspace(1)* %out0, double addrspace(1)* %out1) #0 { 7 store float 0.0, float addrspace(1)* %out0 8 store double 0.0, double addrspace(1)* %out1 9 ret void 10 } 11 12 ; GCN-LABEL: {{^}}test_default_vi: 13 ; GCN: FloatMode: 192 14 ; GCN: IeeeMode: 1 15 define amdgpu_kernel void @test_default_vi(float addrspace(1)* %out0, double addrspace(1)* %out1) #1 { 16 store float 0.0, float addrspace(1)* %out0 17 store double 0.0, double addrspace(1)* %out1 18 ret void 19 } 20 21 ; GCN-LABEL: {{^}}test_f64_denormals: 22 ; GCN: FloatMode: 192 23 ; GCN: IeeeMode: 1 24 define amdgpu_kernel void @test_f64_denormals(float addrspace(1)* %out0, double addrspace(1)* %out1) #2 { 25 store float 0.0, float addrspace(1)* %out0 26 store double 0.0, double addrspace(1)* %out1 27 ret void 28 } 29 30 ; GCN-LABEL: {{^}}test_f32_denormals: 31 ; GCNL: FloatMode: 48 32 ; GCN: IeeeMode: 1 33 define amdgpu_kernel void @test_f32_denormals(float addrspace(1)* %out0, double addrspace(1)* %out1) #3 { 34 store float 0.0, float addrspace(1)* %out0 35 store double 0.0, double addrspace(1)* %out1 36 ret void 37 } 38 39 ; GCN-LABEL: {{^}}test_f32_f64_denormals: 40 ; GCN: FloatMode: 240 41 ; GCN: IeeeMode: 1 42 define amdgpu_kernel void @test_f32_f64_denormals(float addrspace(1)* %out0, double addrspace(1)* %out1) #4 { 43 store float 0.0, float addrspace(1)* %out0 44 store double 0.0, double addrspace(1)* %out1 45 ret void 46 } 47 48 ; GCN-LABEL: {{^}}test_no_denormals 49 ; GCN: FloatMode: 0 50 ; GCN: IeeeMode: 1 51 define amdgpu_kernel void @test_no_denormals(float addrspace(1)* %out0, double addrspace(1)* %out1) #5 { 52 store float 0.0, float addrspace(1)* %out0 53 store double 0.0, double addrspace(1)* %out1 54 ret void 55 } 56 57 ; GCN-LABEL: {{^}}test_f16_f64_denormals: 58 ; GCN: FloatMode: 192 59 ; GCN: IeeeMode: 1 60 define amdgpu_kernel void @test_f16_f64_denormals(half addrspace(1)* %out0, double addrspace(1)* %out1) #6 { 61 store half 0.0, half addrspace(1)* %out0 62 store double 0.0, double addrspace(1)* %out1 63 ret void 64 } 65 66 ; GCN-LABEL: {{^}}test_no_f16_f64_denormals: 67 ; GCN: FloatMode: 0 68 ; GCN: IeeeMode: 1 69 define amdgpu_kernel void @test_no_f16_f64_denormals(half addrspace(1)* %out0, double addrspace(1)* %out1) #7 { 70 store half 0.0, half addrspace(1)* %out0 71 store double 0.0, double addrspace(1)* %out1 72 ret void 73 } 74 75 ; GCN-LABEL: {{^}}test_f32_f16_f64_denormals: 76 ; GCN: FloatMode: 240 77 ; GCN: IeeeMode: 1 78 define amdgpu_kernel void @test_f32_f16_f64_denormals(half addrspace(1)* %out0, float addrspace(1)* %out1, double addrspace(1)* %out2) #8 { 79 store half 0.0, half addrspace(1)* %out0 80 store float 0.0, float addrspace(1)* %out1 81 store double 0.0, double addrspace(1)* %out2 82 ret void 83 } 84 85 ; GCN-LABEL: {{^}}kill_gs_const: 86 ; GCN: IeeeMode: 0 87 define amdgpu_gs void @kill_gs_const() { 88 main_body: 89 %0 = icmp ule i32 0, 3 90 %1 = select i1 %0, float 1.000000e+00, float -1.000000e+00 91 call void @llvm.AMDGPU.kill(float %1) 92 %2 = icmp ule i32 3, 0 93 %3 = select i1 %2, float 1.000000e+00, float -1.000000e+00 94 call void @llvm.AMDGPU.kill(float %3) 95 ret void 96 } 97 98 ; GCN-LABEL: {{^}}kill_vcc_implicit_def: 99 ; GCN: IeeeMode: 0 100 define amdgpu_ps float @kill_vcc_implicit_def([6 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* byval, [17 x <4 x i32>] addrspace(2)* byval, [34 x <8 x i32>] addrspace(2)* byval, float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, i32, float, float) { 101 entry: 102 %tmp0 = fcmp olt float %13, 0.0 103 call void @llvm.AMDGPU.kill(float %14) 104 %tmp1 = select i1 %tmp0, float 1.0, float 0.0 105 ret float %tmp1 106 } 107 108 declare void @llvm.AMDGPU.kill(float) 109 110 attributes #0 = { nounwind "target-cpu"="tahiti" } 111 attributes #1 = { nounwind "target-cpu"="fiji" } 112 attributes #2 = { nounwind "target-features"="+fp64-denormals" } 113 attributes #3 = { nounwind "target-features"="+fp32-denormals" } 114 attributes #4 = { nounwind "target-features"="+fp32-denormals,+fp64-denormals" } 115 attributes #5 = { nounwind "target-features"="-fp32-denormals,-fp64-fp16-denormals" } 116 attributes #6 = { nounwind "target-features"="+fp64-fp16-denormals" } 117 attributes #7 = { nounwind "target-features"="-fp64-fp16-denormals" } 118 attributes #8 = { nounwind "target-features"="+fp32-denormals,+fp64-fp16-denormals" } 119
