1 ; RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI %s 2 3 declare float @llvm.AMDGPU.cvt.f32.ubyte0(i32) nounwind readnone 4 declare float @llvm.AMDGPU.cvt.f32.ubyte1(i32) nounwind readnone 5 declare float @llvm.AMDGPU.cvt.f32.ubyte2(i32) nounwind readnone 6 declare float @llvm.AMDGPU.cvt.f32.ubyte3(i32) nounwind readnone 7 8 ; SI-LABEL: @test_unpack_byte0_to_float: 9 ; SI: V_CVT_F32_UBYTE0 10 define void @test_unpack_byte0_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind { 11 %val = load i32 addrspace(1)* %in, align 4 12 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte0(i32 %val) nounwind readnone 13 store float %cvt, float addrspace(1)* %out, align 4 14 ret void 15 } 16 17 ; SI-LABEL: @test_unpack_byte1_to_float: 18 ; SI: V_CVT_F32_UBYTE1 19 define void @test_unpack_byte1_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind { 20 %val = load i32 addrspace(1)* %in, align 4 21 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte1(i32 %val) nounwind readnone 22 store float %cvt, float addrspace(1)* %out, align 4 23 ret void 24 } 25 26 ; SI-LABEL: @test_unpack_byte2_to_float: 27 ; SI: V_CVT_F32_UBYTE2 28 define void @test_unpack_byte2_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind { 29 %val = load i32 addrspace(1)* %in, align 4 30 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte2(i32 %val) nounwind readnone 31 store float %cvt, float addrspace(1)* %out, align 4 32 ret void 33 } 34 35 ; SI-LABEL: @test_unpack_byte3_to_float: 36 ; SI: V_CVT_F32_UBYTE3 37 define void @test_unpack_byte3_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind { 38 %val = load i32 addrspace(1)* %in, align 4 39 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte3(i32 %val) nounwind readnone 40 store float %cvt, float addrspace(1)* %out, align 4 41 ret void 42 } 43
