• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: opt < %s -S -basicaa -memcpyopt | FileCheck %s
2target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64"
3
4declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i1) nounwind
5declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i1) nounwind
6
7; The resulting memset is only 4-byte aligned, despite containing
8; a 16-byte aligned store in the middle.
9
10define void @foo(i32* %p) {
11; CHECK-LABEL: @foo(
12; CHECK: call void @llvm.memset.p0i8.i64(i8* align 4 {{.*}}, i8 0, i64 16, i1 false)
13  %a0 = getelementptr i32, i32* %p, i64 0
14  store i32 0, i32* %a0, align 4
15  %a1 = getelementptr i32, i32* %p, i64 1
16  store i32 0, i32* %a1, align 16
17  %a2 = getelementptr i32, i32* %p, i64 2
18  store i32 0, i32* %a2, align 4
19  %a3 = getelementptr i32, i32* %p, i64 3
20  store i32 0, i32* %a3, align 4
21  ret void
22}
23
24; Replacing %a8 with %a4 in the memset requires boosting the alignment of %a4.
25
26define void @bar() {
27; CHECK-LABEL: @bar(
28; CHECK: %a4 = alloca i32, align 8
29; CHECK-NOT: memcpy
30  %a4 = alloca i32, align 4
31  %a8 = alloca i32, align 8
32  %a8.cast = bitcast i32* %a8 to i8*
33  %a4.cast = bitcast i32* %a4 to i8*
34  call void @llvm.memset.p0i8.i64(i8* align 8 %a8.cast, i8 0, i64 4, i1 false)
35  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %a4.cast, i8* align 4 %a8.cast, i64 4, i1 false)
36  ret void
37}
38