| Kyle Butt | 8c0314c | 2017-09-09 00:37:56 +0000 | [diff] [blame^] | 1 | ; RUN: llc -O3 -o - %s | FileCheck %s |
| 2 | target datalayout = "e-m:e-i64:64-n32:64" |
| 3 | target triple = "powerpc64le-unknown-linux-gnu" |
| 4 | |
| 5 | %class1 = type { %union1 } |
| 6 | %union1 = type { i64, [24 x i8] } |
| 7 | %class2 = type { %class3 } |
| 8 | %class3 = type { %class4 } |
| 9 | %class4 = type { %class5, i64, %union.anon } |
| 10 | %class5 = type { i8* } |
| 11 | %union.anon = type { i64, [8 x i8] } |
| 12 | |
| 13 | @ext = external global %"class1", align 8 |
| 14 | |
| 15 | ; We can't select lxv for this because even though we're accessing an offset of |
| 16 | ; 16 from the stack slot, the stack slot is only guaranteed to be 8-byte |
| 17 | ; aligned. When the frame is finalized it is converted to lxv (frame-reg) + |
| 18 | ; (offset + 16). Because offset isn't guaranteed to be 16-byte aligned, we may |
| 19 | ; end up needing to translate the lxv instruction to lxvx |
| 20 | ; CHECK-LABEL: unaligned_slot: |
| 21 | ; CHECK-NOT: lxv {{[0-9]+}}, {{[-0-9]+}}({{[0-9]+}}) |
| 22 | ; CHECK: blr |
| 23 | define void @unaligned_slot() #0 { |
| 24 | %1 = alloca %class2, align 8 |
| 25 | %2 = getelementptr inbounds %class2, %class2* %1, i64 0, i32 0, i32 0, i32 2 |
| 26 | %3 = bitcast %union.anon* %2 to i8* |
| 27 | call void @llvm.memcpy.p0i8.p0i8.i64(i8* nonnull getelementptr inbounds (%class1, %class1* @ext, i64 0, i32 0, i32 1, i64 8), i8* nonnull %3, i64 16, i32 8, i1 false) #2 |
| 28 | ret void |
| 29 | } |
| 30 | ; CHECK-LABEL: aligned_slot: |
| 31 | ; CHECK: lxv {{[0-9]+}}, {{[-0-9]+}}({{[0-9]+}}) |
| 32 | ; CHECK: blr |
| 33 | define void @aligned_slot() #0 { |
| 34 | %1 = alloca %class2, align 16 |
| 35 | %2 = getelementptr inbounds %class2, %class2* %1, i64 0, i32 0, i32 0, i32 2 |
| 36 | %3 = bitcast %union.anon* %2 to i8* |
| 37 | call void @llvm.memcpy.p0i8.p0i8.i64(i8* nonnull getelementptr inbounds (%class1, %class1* @ext, i64 0, i32 0, i32 1, i64 8), i8* nonnull %3, i64 16, i32 8, i1 false) #2 |
| 38 | ret void |
| 39 | } |
| 40 | |
| 41 | ; Function Attrs: argmemonly nounwind |
| 42 | declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1) #1 |
| 43 | |
| 44 | attributes #0 = { nounwind "target-cpu"="pwr9" "target-features"="+altivec,+bpermd,+crypto,+direct-move,+extdiv,+htm,+power8-vector,+power9-vector,+vsx,-qpx" "unsafe-fp-math"="false" "use-soft-float"="false" } |
| 45 | attributes #1 = { argmemonly nounwind } |
| 46 | attributes #2 = { nounwind } |