1; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=core2 | FileCheck %s -check-prefix=LINUX
2; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core2 | FileCheck %s -check-prefix=DARWIN
3
4declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
5declare void @llvm.memcpy.p256i8.p256i8.i64(i8 addrspace(256)* nocapture, i8 addrspace(256)* nocapture, i64, i32, i1) nounwind
6
7
8; Variable memcpy's should lower to calls.
9define i8* @test1(i8* %a, i8* %b, i64 %n) nounwind {
10entry:
11	tail call void @llvm.memcpy.p0i8.p0i8.i64( i8* %a, i8* %b, i64 %n, i32 1, i1 0 )
12	ret i8* %a
13
14; LINUX-LABEL: test1:
15; LINUX: memcpy
16}
17
18; Variable memcpy's should lower to calls.
19define i8* @test2(i64* %a, i64* %b, i64 %n) nounwind {
20entry:
21	%tmp14 = bitcast i64* %a to i8*
22	%tmp25 = bitcast i64* %b to i8*
23	tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %tmp14, i8* %tmp25, i64 %n, i32 8, i1 0 )
24	ret i8* %tmp14
25
26; LINUX-LABEL: test2:
27; LINUX: memcpy
28}
29
30; Large constant memcpy's should lower to a call when optimizing for size.
31; PR6623
32
33; On the other hand, Darwin's definition of -Os is optimizing for size without
34; hurting performance so it should just ignore optsize when expanding memcpy.
35; rdar://8821501
36define void @test3(i8* nocapture %A, i8* nocapture %B) nounwind optsize noredzone {
37entry:
38  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i32 1, i1 false)
39  ret void
40; LINUX-LABEL: test3:
41; LINUX: memcpy
42
43; DARWIN-LABEL: test3:
44; DARWIN-NOT: memcpy
45; DARWIN: movq
46; DARWIN: movq
47; DARWIN: movq
48; DARWIN: movq
49; DARWIN: movq
50; DARWIN: movq
51; DARWIN: movq
52; DARWIN: movq
53; DARWIN: movq
54; DARWIN: movq
55; DARWIN: movq
56; DARWIN: movq
57; DARWIN: movq
58; DARWIN: movq
59; DARWIN: movq
60; DARWIN: movq
61}
62
63define void @test3_minsize(i8* nocapture %A, i8* nocapture %B) nounwind minsize noredzone {
64  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i32 1, i1 false)
65  ret void
66; LINUX-LABEL: test3_minsize:
67; LINUX: memcpy
68
69; DARWIN-LABEL: test3_minsize:
70; DARWIN: memcpy
71}
72
73define void @test3_minsize_optsize(i8* nocapture %A, i8* nocapture %B) nounwind optsize minsize noredzone {
74  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i32 1, i1 false)
75  ret void
76; LINUX-LABEL: test3_minsize_optsize:
77; LINUX: memcpy
78
79; DARWIN-LABEL: test3_minsize_optsize:
80; DARWIN: memcpy
81}
82
83; Large constant memcpy's should be inlined when not optimizing for size.
84define void @test4(i8* nocapture %A, i8* nocapture %B) nounwind noredzone {
85entry:
86  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i32 1, i1 false)
87  ret void
88; LINUX-LABEL: test4:
89; LINUX: movq
90; LINUX: movq
91; LINUX: movq
92; LINUX: movq
93; LINUX: movq
94; LINUX: movq
95; LINUX: movq
96; LINUX: movq
97; LINUX: movq
98; LINUX: movq
99; LINUX: movq
100; LINUX: movq
101}
102
103
104@.str = private unnamed_addr constant [30 x i8] c"\00aaaaaaaaaaaaaaaaaaaaaaaaaaaa\00", align 1
105
106define void @test5(i8* nocapture %C) nounwind uwtable ssp {
107entry:
108  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %C, i8* getelementptr inbounds ([30 x i8], [30 x i8]* @.str, i64 0, i64 0), i64 16, i32 1, i1 false)
109  ret void
110
111; DARWIN-LABEL: test5:
112; DARWIN: movabsq	$7016996765293437281
113; DARWIN: movabsq	$7016996765293437184
114}
115
116
117; PR14896
118@.str2 = private unnamed_addr constant [2 x i8] c"x\00", align 1
119
120define void @test6() nounwind uwtable {
121entry:
122; DARWIN: test6
123; DARWIN: movw $0, 8
124; DARWIN: movq $120, 0
125  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* null, i8* getelementptr inbounds ([2 x i8], [2 x i8]* @.str2, i64 0, i64 0), i64 10, i32 1, i1 false)
126  ret void
127}
128
129define void @PR15348(i8* %a, i8* %b) {
130; Ensure that alignment of '0' in an @llvm.memcpy intrinsic results in
131; unaligned loads and stores.
132; LINUX: PR15348
133; LINUX: movb
134; LINUX: movb
135; LINUX: movq
136; LINUX: movq
137; LINUX: movq
138; LINUX: movq
139  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* %b, i64 17, i32 0, i1 false)
140  ret void
141}
142
143; Memcpys from / to address space 256 should be lowered to appropriate loads /
144; stores if small enough.
145define void @addrspace256(i8 addrspace(256)* %a, i8 addrspace(256)* %b) nounwind {
146  tail call void @llvm.memcpy.p256i8.p256i8.i64(i8 addrspace(256)* %a, i8 addrspace(256)* %b, i64 16, i32 8, i1 false)
147  ret void
148; LINUX-LABEL: addrspace256:
149; LINUX: movq %gs:
150; LINUX: movq %gs:
151; LINUX: movq {{.*}}, %gs:
152; LINUX: movq {{.*}}, %gs:
153}
154