Home | History | Annotate | Download | only in AMDGPU
      1 ; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
      2 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
      3 
      4 ; This test just checks that the compiler doesn't crash.
      5 
      6 declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
      7 
      8 ; FUNC-LABEL: {{^}}v32i8_to_v8i32:
      9 ; SI: s_endpgm
     10 define void @v32i8_to_v8i32(<32 x i8> addrspace(2)* inreg) #0 {
     11 entry:
     12   %1 = load <32 x i8>, <32 x i8> addrspace(2)* %0
     13   %2 = bitcast <32 x i8> %1 to <8 x i32>
     14   %3 = extractelement <8 x i32> %2, i32 1
     15   %4 = icmp ne i32 %3, 0
     16   %5 = select i1 %4, float 0.0, float 1.0
     17   call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %5, float %5, float %5, float %5)
     18   ret void
     19 }
     20 
     21 ; FUNC-LABEL: {{^}}i8ptr_v16i8ptr:
     22 ; SI: s_endpgm
     23 define void @i8ptr_v16i8ptr(<16 x i8> addrspace(1)* %out, i8 addrspace(1)* %in) {
     24 entry:
     25   %0 = bitcast i8 addrspace(1)* %in to <16 x i8> addrspace(1)*
     26   %1 = load <16 x i8>, <16 x i8> addrspace(1)* %0
     27   store <16 x i8> %1, <16 x i8> addrspace(1)* %out
     28   ret void
     29 }
     30 
     31 define void @f32_to_v2i16(<2 x i16> addrspace(1)* %out, float addrspace(1)* %in) nounwind {
     32   %load = load float, float addrspace(1)* %in, align 4
     33   %bc = bitcast float %load to <2 x i16>
     34   store <2 x i16> %bc, <2 x i16> addrspace(1)* %out, align 4
     35   ret void
     36 }
     37 
     38 define void @v2i16_to_f32(float addrspace(1)* %out, <2 x i16> addrspace(1)* %in) nounwind {
     39   %load = load <2 x i16>, <2 x i16> addrspace(1)* %in, align 4
     40   %bc = bitcast <2 x i16> %load to float
     41   store float %bc, float addrspace(1)* %out, align 4
     42   ret void
     43 }
     44 
     45 define void @v4i8_to_i32(i32 addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
     46   %load = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
     47   %bc = bitcast <4 x i8> %load to i32
     48   store i32 %bc, i32 addrspace(1)* %out, align 4
     49   ret void
     50 }
     51 
     52 define void @i32_to_v4i8(<4 x i8> addrspace(1)* %out, i32 addrspace(1)* %in) nounwind {
     53   %load = load i32, i32 addrspace(1)* %in, align 4
     54   %bc = bitcast i32 %load to <4 x i8>
     55   store <4 x i8> %bc, <4 x i8> addrspace(1)* %out, align 4
     56   ret void
     57 }
     58 
     59 ; FUNC-LABEL: {{^}}bitcast_v2i32_to_f64:
     60 ; SI: s_endpgm
     61 define void @bitcast_v2i32_to_f64(double addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
     62   %val = load <2 x i32>, <2 x i32> addrspace(1)* %in, align 8
     63   %add = add <2 x i32> %val, <i32 4, i32 9>
     64   %bc = bitcast <2 x i32> %add to double
     65   store double %bc, double addrspace(1)* %out, align 8
     66   ret void
     67 }
     68 
     69 ; FUNC-LABEL: {{^}}bitcast_f64_to_v2i32:
     70 ; SI: s_endpgm
     71 define void @bitcast_f64_to_v2i32(<2 x i32> addrspace(1)* %out, double addrspace(1)* %in) {
     72   %val = load double, double addrspace(1)* %in, align 8
     73   %add = fadd double %val, 4.0
     74   %bc = bitcast double %add to <2 x i32>
     75   store <2 x i32> %bc, <2 x i32> addrspace(1)* %out, align 8
     76   ret void
     77 }
     78 
     79 attributes #0 = { "ShaderType"="0" }
     80