1 ; RUN: llc -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI %s
2 ; RUN: llc -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=SI %s
4 declare float @llvm.AMDGPU.cvt.f32.ubyte0(i32) nounwind readnone
5 declare float @llvm.AMDGPU.cvt.f32.ubyte1(i32) nounwind readnone
6 declare float @llvm.AMDGPU.cvt.f32.ubyte2(i32) nounwind readnone
7 declare float @llvm.AMDGPU.cvt.f32.ubyte3(i32) nounwind readnone
9 ; SI-LABEL: {{^}}test_unpack_byte0_to_float:
10 ; SI: v_cvt_f32_ubyte0
11 define void @test_unpack_byte0_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
12 %val = load i32, i32 addrspace(1)* %in, align 4
13 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte0(i32 %val) nounwind readnone
14 store float %cvt, float addrspace(1)* %out, align 4
18 ; SI-LABEL: {{^}}test_unpack_byte1_to_float:
19 ; SI: v_cvt_f32_ubyte1
20 define void @test_unpack_byte1_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
21 %val = load i32, i32 addrspace(1)* %in, align 4
22 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte1(i32 %val) nounwind readnone
23 store float %cvt, float addrspace(1)* %out, align 4
27 ; SI-LABEL: {{^}}test_unpack_byte2_to_float:
28 ; SI: v_cvt_f32_ubyte2
29 define void @test_unpack_byte2_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
30 %val = load i32, i32 addrspace(1)* %in, align 4
31 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte2(i32 %val) nounwind readnone
32 store float %cvt, float addrspace(1)* %out, align 4
36 ; SI-LABEL: {{^}}test_unpack_byte3_to_float:
37 ; SI: v_cvt_f32_ubyte3
38 define void @test_unpack_byte3_to_float(float addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
39 %val = load i32, i32 addrspace(1)* %in, align 4
40 %cvt = call float @llvm.AMDGPU.cvt.f32.ubyte3(i32 %val) nounwind readnone
41 store float %cvt, float addrspace(1)* %out, align 4