diff options
Diffstat (limited to 'test/CodeGen/AMDGPU/schedule-vs-if-nested-loop-failure.ll')
| -rw-r--r-- | test/CodeGen/AMDGPU/schedule-vs-if-nested-loop-failure.ll | 51 |
1 files changed, 23 insertions, 28 deletions
diff --git a/test/CodeGen/AMDGPU/schedule-vs-if-nested-loop-failure.ll b/test/CodeGen/AMDGPU/schedule-vs-if-nested-loop-failure.ll index e4b16c0a165f..9b490bb3a731 100644 --- a/test/CodeGen/AMDGPU/schedule-vs-if-nested-loop-failure.ll +++ b/test/CodeGen/AMDGPU/schedule-vs-if-nested-loop-failure.ll @@ -1,13 +1,10 @@ -; XFAIL: * -; REQUIRES: asserts -; RUN: llc -O0 -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck %s -check-prefix=SI -; RUN: llc -O0 -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck %s -check-prefix=SI +; RUN: llc -O0 -march=amdgcn -verify-machineinstrs -mattr=+vgpr-spilling < %s | FileCheck -check-prefix=GCN %s +; RUN: llc -O0 -march=amdgcn -mcpu=tonga -verify-machineinstrs -mattr=+vgpr-spilling < %s | FileCheck -check-prefix=GCN %s -declare void @llvm.AMDGPU.barrier.local() nounwind convergent +declare void @llvm.amdgcn.s.barrier() nounwind convergent - -; SI-LABEL: {{^}}main( -define void @main(<4 x float> inreg %reg0, <4 x float> inreg %reg1) #0 { +; GCN-LABEL: {{^}}main: +define amdgpu_vs void @main(<4 x float> inreg %reg0, <4 x float> inreg %reg1) { main_body: %0 = extractelement <4 x float> %reg1, i32 0 %1 = extractelement <4 x float> %reg1, i32 2 @@ -39,63 +36,63 @@ ENDIF: ; preds = %main_body, %Flow2 %temp3.0 = phi float [ 0.000000e+00, %main_body ], [ %101, %Flow2 ] %15 = extractelement <4 x float> %reg1, i32 1 %16 = extractelement <4 x float> %reg1, i32 3 - %17 = load <4 x float>, <4 x float> addrspace(9)* null + %17 = load <4 x float>, <4 x float> addrspace(2)* null %18 = extractelement <4 x float> %17, i32 0 %19 = fmul float %18, %0 - %20 = load <4 x float>, <4 x float> addrspace(9)* null + %20 = load <4 x float>, <4 x float> addrspace(2)* null %21 = extractelement <4 x float> %20, i32 1 %22 = fmul float %21, %0 - %23 = load <4 x float>, <4 x float> addrspace(9)* null + %23 = load <4 x float>, <4 x float> addrspace(2)* null %24 = extractelement <4 x float> %23, i32 2 %25 = fmul float %24, %0 - %26 = load <4 x float>, <4 x float> addrspace(9)* null + %26 = load <4 x float>, <4 x float> addrspace(2)* null %27 = extractelement <4 x float> %26, i32 3 %28 = fmul float %27, %0 - %29 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 1) + %29 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 1) %30 = extractelement <4 x float> %29, i32 0 %31 = fmul float %30, %15 %32 = fadd float %31, %19 - %33 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 1) + %33 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 1) %34 = extractelement <4 x float> %33, i32 1 %35 = fmul float %34, %15 %36 = fadd float %35, %22 - %37 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 1) + %37 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 1) %38 = extractelement <4 x float> %37, i32 2 %39 = fmul float %38, %15 %40 = fadd float %39, %25 - %41 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 1) + %41 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 1) %42 = extractelement <4 x float> %41, i32 3 %43 = fmul float %42, %15 %44 = fadd float %43, %28 - %45 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 2) + %45 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 2) %46 = extractelement <4 x float> %45, i32 0 %47 = fmul float %46, %1 %48 = fadd float %47, %32 - %49 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 2) + %49 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 2) %50 = extractelement <4 x float> %49, i32 1 %51 = fmul float %50, %1 %52 = fadd float %51, %36 - %53 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 2) + %53 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 2) %54 = extractelement <4 x float> %53, i32 2 %55 = fmul float %54, %1 %56 = fadd float %55, %40 - %57 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 2) + %57 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 2) %58 = extractelement <4 x float> %57, i32 3 %59 = fmul float %58, %1 %60 = fadd float %59, %44 - %61 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 3) + %61 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 3) %62 = extractelement <4 x float> %61, i32 0 %63 = fmul float %62, %16 %64 = fadd float %63, %48 - %65 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 3) + %65 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 3) %66 = extractelement <4 x float> %65, i32 1 %67 = fmul float %66, %16 %68 = fadd float %67, %52 - %69 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 3) + %69 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 3) %70 = extractelement <4 x float> %69, i32 2 %71 = fmul float %70, %16 %72 = fadd float %71, %56 - %73 = load <4 x float>, <4 x float> addrspace(9)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(9)* null, i64 0, i32 3) + %73 = load <4 x float>, <4 x float> addrspace(2)* getelementptr ([1024 x <4 x float>], [1024 x <4 x float>] addrspace(2)* null, i64 0, i32 3) %74 = extractelement <4 x float> %73, i32 3 %75 = fmul float %74, %16 %76 = fadd float %75, %60 @@ -103,12 +100,12 @@ ENDIF: ; preds = %main_body, %Flow2 %78 = insertelement <4 x float> %77, float %68, i32 1 %79 = insertelement <4 x float> %78, float %72, i32 2 %80 = insertelement <4 x float> %79, float %76, i32 3 - call void @llvm.AMDGPU.barrier.local() + call void @llvm.amdgcn.s.barrier() %81 = insertelement <4 x float> undef, float %temp.0, i32 0 %82 = insertelement <4 x float> %81, float %temp1.0, i32 1 %83 = insertelement <4 x float> %82, float %temp2.0, i32 2 %84 = insertelement <4 x float> %83, float %temp3.0, i32 3 - call void @llvm.AMDGPU.barrier.local() + call void @llvm.amdgcn.s.barrier() ret void LOOP: ; preds = %main_body, %Flow @@ -159,5 +156,3 @@ ENDIF19: ; preds = %ENDIF16 %115 = fadd float %temp4.0, 1.000000e+00 br label %Flow1 } - -attributes #0 = { "ShaderType"="1" } |