1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt < %s -S -basic-aa -memcpyopt -verify-memoryssa | FileCheck %s
3 target 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"
5 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i1) nounwind
6 declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i1) nounwind
8 ; The resulting memset is only 4-byte aligned, despite containing
9 ; a 16-byte aligned store in the middle.
11 define void @foo(i32* %p) {
13 ; CHECK-NEXT: [[A0:%.*]] = getelementptr i32, i32* [[P:%.*]], i64 0
14 ; CHECK-NEXT: [[A1:%.*]] = getelementptr i32, i32* [[P]], i64 1
15 ; CHECK-NEXT: [[A2:%.*]] = getelementptr i32, i32* [[P]], i64 2
16 ; CHECK-NEXT: [[A3:%.*]] = getelementptr i32, i32* [[P]], i64 3
17 ; CHECK-NEXT: [[TMP1:%.*]] = bitcast i32* [[A0]] to i8*
18 ; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* align 4 [[TMP1]], i8 0, i64 16, i1 false)
19 ; CHECK-NEXT: ret void
21 %a0 = getelementptr i32, i32* %p, i64 0
22 store i32 0, i32* %a0, align 4
23 %a1 = getelementptr i32, i32* %p, i64 1
24 store i32 0, i32* %a1, align 16
25 %a2 = getelementptr i32, i32* %p, i64 2
26 store i32 0, i32* %a2, align 4
27 %a3 = getelementptr i32, i32* %p, i64 3
28 store i32 0, i32* %a3, align 4
32 ; Replacing %a8 with %a4 in the memset requires boosting the alignment of %a4.
36 ; CHECK-NEXT: [[A4:%.*]] = alloca i32, align 8
37 ; CHECK-NEXT: [[A8:%.*]] = alloca i32, align 8
38 ; CHECK-NEXT: [[A8_CAST:%.*]] = bitcast i32* [[A8]] to i8*
39 ; CHECK-NEXT: [[A4_CAST:%.*]] = bitcast i32* [[A4]] to i8*
40 ; CHECK-NEXT: [[A41:%.*]] = bitcast i32* [[A4]] to i8*
41 ; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[A41]], i8 0, i64 4, i1 false)
42 ; CHECK-NEXT: ret void
44 %a4 = alloca i32, align 4
45 %a8 = alloca i32, align 8
46 %a8.cast = bitcast i32* %a8 to i8*
47 %a4.cast = bitcast i32* %a4 to i8*
48 call void @llvm.memset.p0i8.i64(i8* align 8 %a8.cast, i8 0, i64 4, i1 false)
49 call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %a4.cast, i8* align 4 %a8.cast, i64 4, i1 false)