Tom Stellard | 2f7cdda | 2013-08-06 23:08:28 +0000 | [diff] [blame] | 1 | ; RUN: llc < %s -march=r600 -mcpu=SI | FileCheck %s |
| 2 | |
| 3 | ; This test checks that no VGPR to SGPR copies are created by the register |
| 4 | ; allocator. |
| 5 | ; CHECK: @main |
| 6 | ; CHECK: S_BUFFER_LOAD_DWORD [[DST:SGPR[0-9]]], {{[SGPR_[0-9]+}}, 0 |
| 7 | ; CHECK: V_MOV_B32_e32 VGPR{{[0-9]}}, [[DST]] |
| 8 | |
| 9 | define void @main(<16 x i8> addrspace(2)* inreg, <16 x i8> addrspace(2)* inreg, <32 x i8> addrspace(2)* 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, float, float, float) #0 { |
| 10 | main_body: |
| 11 | %20 = getelementptr <16 x i8> addrspace(2)* %0, i32 0 |
| 12 | %21 = load <16 x i8> addrspace(2)* %20, !tbaa !0 |
| 13 | %22 = call float @llvm.SI.load.const(<16 x i8> %21, i32 0) |
| 14 | %23 = call float @llvm.SI.load.const(<16 x i8> %21, i32 16) |
| 15 | %24 = call float @llvm.SI.load.const(<16 x i8> %21, i32 32) |
| 16 | %25 = fptosi float %23 to i32 |
| 17 | %26 = icmp ne i32 %25, 0 |
| 18 | br i1 %26, label %ENDIF, label %ELSE |
| 19 | |
| 20 | ELSE: ; preds = %main_body |
| 21 | %27 = fsub float -0.000000e+00, %22 |
| 22 | br label %ENDIF |
| 23 | |
| 24 | ENDIF: ; preds = %main_body, %ELSE |
| 25 | %temp.0 = phi float [ %27, %ELSE ], [ %22, %main_body ] |
| 26 | %28 = fadd float %temp.0, %24 |
| 27 | call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %28, float %28, float 0.000000e+00, float 1.000000e+00) |
| 28 | ret void |
| 29 | } |
| 30 | |
| 31 | ; We just want ot make sure the program doesn't crash |
| 32 | ; CHECK: @loop |
| 33 | |
| 34 | define void @loop(<16 x i8> addrspace(2)* inreg, <16 x i8> addrspace(2)* inreg, <32 x i8> addrspace(2)* 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, float, float, float) #0 { |
| 35 | main_body: |
| 36 | %20 = getelementptr <16 x i8> addrspace(2)* %0, i32 0 |
| 37 | %21 = load <16 x i8> addrspace(2)* %20, !tbaa !0 |
| 38 | %22 = call float @llvm.SI.load.const(<16 x i8> %21, i32 0) |
| 39 | %23 = call float @llvm.SI.load.const(<16 x i8> %21, i32 4) |
| 40 | %24 = call float @llvm.SI.load.const(<16 x i8> %21, i32 8) |
| 41 | %25 = call float @llvm.SI.load.const(<16 x i8> %21, i32 12) |
| 42 | %26 = fptosi float %25 to i32 |
| 43 | %27 = bitcast i32 %26 to float |
| 44 | %28 = bitcast float %27 to i32 |
| 45 | br label %LOOP |
| 46 | |
| 47 | LOOP: ; preds = %ENDIF, %main_body |
| 48 | %temp4.0 = phi float [ %22, %main_body ], [ %temp5.0, %ENDIF ] |
| 49 | %temp5.0 = phi float [ %23, %main_body ], [ %temp6.0, %ENDIF ] |
| 50 | %temp6.0 = phi float [ %24, %main_body ], [ %temp4.0, %ENDIF ] |
| 51 | %temp8.0 = phi float [ 0.000000e+00, %main_body ], [ %37, %ENDIF ] |
| 52 | %29 = bitcast float %temp8.0 to i32 |
| 53 | %30 = icmp sge i32 %29, %28 |
| 54 | %31 = sext i1 %30 to i32 |
| 55 | %32 = bitcast i32 %31 to float |
| 56 | %33 = bitcast float %32 to i32 |
| 57 | %34 = icmp ne i32 %33, 0 |
| 58 | br i1 %34, label %IF, label %ENDIF |
| 59 | |
| 60 | IF: ; preds = %LOOP |
| 61 | call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %temp4.0, float %temp5.0, float %temp6.0, float 1.000000e+00) |
| 62 | ret void |
| 63 | |
| 64 | ENDIF: ; preds = %LOOP |
| 65 | %35 = bitcast float %temp8.0 to i32 |
| 66 | %36 = add i32 %35, 1 |
| 67 | %37 = bitcast i32 %36 to float |
| 68 | br label %LOOP |
| 69 | } |
| 70 | |
| 71 | ; Function Attrs: nounwind readnone |
| 72 | declare float @llvm.SI.load.const(<16 x i8>, i32) #1 |
| 73 | |
| 74 | ; Function Attrs: readonly |
| 75 | declare float @fabs(float) #2 |
| 76 | |
| 77 | declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float) |
| 78 | |
| 79 | attributes #0 = { "ShaderType"="0" } |
| 80 | attributes #1 = { nounwind readnone } |
| 81 | attributes #2 = { readonly } |
| 82 | |
| 83 | !0 = metadata !{metadata !"const", null, i32 1} |
| 84 | |