1; MemCpy optimizations should take place even in presence of invariant.start
2; RUN: opt < %s -basicaa -memcpyopt -dse -S | FileCheck %s
3
4target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
5
6target triple = "i686-apple-darwin9"
7
8%0 = type { x86_fp80, x86_fp80 }
9declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32,
10i1) nounwind
11declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly,
12i64, i32, i1)
13declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1)
14
15declare {}* @llvm.invariant.start.p0i8(i64, i8* nocapture) nounwind readonly
16
17; FIXME: The invariant.start does not modify %P.
18; The intermediate alloca and one of the memcpy's should be eliminated, the
19; other should be transformed to a memmove.
20define void @test1(i8* %P, i8* %Q) nounwind  {
21  %memtmp = alloca %0, align 16
22  %R = bitcast %0* %memtmp to i8*
23  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %R, i8* %P, i32 32, i32 16, i1 false)
24  %i = call {}* @llvm.invariant.start.p0i8(i64 32, i8* %P)
25  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %Q, i8* %R, i32 32, i32 16, i1 false)
26  ret void
27; CHECK-LABEL: @test1(
28; CHECK-NEXT: %memtmp = alloca %0, align 16
29; CHECK-NEXT: %R = bitcast %0* %memtmp to i8*
30; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %R, i8* %P, i32 32, i32 16, i1 false)
31; CHECK-NEXT: %i = call {}* @llvm.invariant.start.p0i8(i64 32, i8* %P)
32; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %Q, i8* %R, i32 32, i32 16, i1 false)
33; CHECK-NEXT: ret void
34}
35
36
37; The invariant.start intrinsic does not inhibit tranforming the memcpy to a
38; memset.
39define void @test2(i8* %dst1, i8* %dst2, i8 %c) {
40; CHECK-LABEL: define void @test2(
41; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* %dst1, i8 %c, i64 128, i32 1, i1 false)
42; CHECK-NEXT: %i = call {}* @llvm.invariant.start.p0i8(i64 32, i8* %dst1)
43; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* %dst2, i8 %c, i64 128, i32 8, i1 false)
44; CHECK-NEXT: ret void
45  call void @llvm.memset.p0i8.i64(i8* %dst1, i8 %c, i64 128, i32 1, i1 false)
46  %i = call {}* @llvm.invariant.start.p0i8(i64 32, i8* %dst1)
47  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst2, i8* %dst1, i64 128, i32 8, i1 false)
48  ret void
49}
50