|
| 1 | +; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s |
| 2 | + |
| 3 | +target datalayout = "e-p:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024" |
| 4 | +target triple = "spir" |
| 5 | + |
| 6 | +%struct.B = type { [2 x i32] } |
| 7 | +%struct.A = type { i64, %struct.B } |
| 8 | + |
| 9 | +@__const.foo.b = private unnamed_addr addrspace(2) constant %struct.B { [2 x i32] [i32 1, i32 2] }, align 4 |
| 10 | +@__const.bar.a = private unnamed_addr addrspace(2) constant %struct.A { i64 0, %struct.B { [2 x i32] [i32 1, i32 2] } }, align 8 |
| 11 | + |
| 12 | +; Function Attrs: convergent nounwind |
| 13 | +define spir_func void @foo(%struct.A* noalias sret(%struct.A) %agg.result) #0 { |
| 14 | +entry: |
| 15 | + %b = alloca %struct.B, align 4 |
| 16 | + %0 = bitcast %struct.B* %b to i8* |
| 17 | + call void @llvm.lifetime.start.p0i8(i64 8, i8* %0) #2 |
| 18 | + %1 = bitcast %struct.B* %b to i8* |
| 19 | + call void @llvm.memcpy.p0i8.p2i8.i32(i8* align 4 %1, i8 addrspace(2)* align 4 bitcast (%struct.B addrspace(2)* @__const.foo.b to i8 addrspace(2)*), i32 8, i1 false) |
| 20 | +; CHECK: OpCopyMemorySized %[[#]] %[[#]] %[[#]] Aligned 4 |
| 21 | + %b1 = getelementptr inbounds %struct.A, %struct.A* %agg.result, i32 0, i32 1 |
| 22 | + %2 = bitcast %struct.B* %b1 to i8* |
| 23 | + %3 = bitcast %struct.B* %b to i8* |
| 24 | + call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %2, i8* align 4 %3, i32 8, i1 false), !tbaa.struct !4 |
| 25 | +; CHECK: %[[#PTR1:]] = OpInBoundsPtrAccessChain %[[#]] %[[#]] %[[#]] %[[#]] |
| 26 | +; CHECK: OpCopyMemorySized %[[#PTR1]] %[[#]] %[[#]] Aligned 8 |
| 27 | + %4 = bitcast %struct.B* %b to i8* |
| 28 | + call void @llvm.lifetime.end.p0i8(i64 8, i8* %4) #2 |
| 29 | + ret void |
| 30 | +} |
| 31 | + |
| 32 | +; Function Attrs: argmemonly nounwind |
| 33 | +declare void @llvm.lifetime.start.p0i8(i64, i8* captures(none)) #1 |
| 34 | + |
| 35 | +; Function Attrs: argmemonly nounwind |
| 36 | +declare void @llvm.memcpy.p0i8.p2i8.i32(i8* captures(none) writeonly, i8 addrspace(2)* captures(none) readonly, i32, i1) #1 |
| 37 | + |
| 38 | +; Function Attrs: argmemonly nounwind |
| 39 | +declare void @llvm.memcpy.p0i8.p0i8.i32(i8* captures(none) writeonly, i8* captures(none) readonly, i32, i1) #1 |
| 40 | + |
| 41 | +; Function Attrs: argmemonly nounwind |
| 42 | +declare void @llvm.lifetime.end.p0i8(i64, i8* captures(none)) #1 |
| 43 | + |
| 44 | +; Function Attrs: convergent nounwind |
| 45 | +define spir_func void @bar(%struct.B* noalias sret(%struct.B) %agg.result) #0 { |
| 46 | +entry: |
| 47 | + %a = alloca %struct.A, align 8 |
| 48 | + %0 = bitcast %struct.A* %a to i8* |
| 49 | + call void @llvm.lifetime.start.p0i8(i64 16, i8* %0) #2 |
| 50 | + %1 = bitcast %struct.A* %a to i8* |
| 51 | + call void @llvm.memcpy.p0i8.p2i8.i32(i8* align 8 %1, i8 addrspace(2)* align 8 bitcast (%struct.A addrspace(2)* @__const.bar.a to i8 addrspace(2)*), i32 16, i1 false) |
| 52 | +; CHECK: OpCopyMemorySized %[[#]] %[[#]] %[[#]] Aligned 8 |
| 53 | + %b = getelementptr inbounds %struct.A, %struct.A* %a, i32 0, i32 1 |
| 54 | + %2 = bitcast %struct.B* %agg.result to i8* |
| 55 | + %3 = bitcast %struct.B* %b to i8* |
| 56 | + call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %2, i8* align 8 %3, i32 8, i1 false), !tbaa.struct !4 |
| 57 | +; CHECK: %[[#PTR2:]] = OpInBoundsPtrAccessChain %[[#]] %[[#]] %[[#]] %[[#]] |
| 58 | +; CHECK: OpCopyMemorySized %[[#]] %[[#PTR2]] %[[#]] Aligned 4 |
| 59 | + %4 = bitcast %struct.A* %a to i8* |
| 60 | + call void @llvm.lifetime.end.p0i8(i64 16, i8* %4) #2 |
| 61 | + ret void |
| 62 | +} |
| 63 | + |
| 64 | +attributes #0 = { convergent nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "denorms-are-zero"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" } |
| 65 | +attributes #1 = { argmemonly nounwind } |
| 66 | +attributes #2 = { nounwind } |
| 67 | + |
| 68 | +!llvm.module.flags = !{!0} |
| 69 | +!opencl.ocl.version = !{!1} |
| 70 | +!opencl.spir.version = !{!2} |
| 71 | +!llvm.ident = !{!3} |
| 72 | + |
| 73 | +!0 = !{i32 1, !"wchar_size", i32 4} |
| 74 | +!1 = !{i32 1, i32 0} |
| 75 | +!2 = !{i32 1, i32 2} |
| 76 | +!3 = !{!"clang version 9.0.0"} |
| 77 | +!4 = !{i64 0, i64 8, !5} |
| 78 | +!5 = !{!6, !6, i64 0} |
| 79 | +!6 = !{!"omnipotent char", !7, i64 0} |
| 80 | +!7 = !{!"Simple C/C++ TBAA"} |
0 commit comments