1; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -tail-dup-placement=0 | FileCheck %s
2
3; Test memcpy, memmove, and memset intrinsics.
4
5target triple = "wasm32-unknown-unknown"
6
7declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i1)
8declare void @llvm.memmove.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i1)
9declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i1)
10
11; Test that return values are optimized.
12
13; CHECK-LABEL: copy_yes:
14; CHECK:      call     $push0=, memcpy, $0, $1, $2{{$}}
15; CHECK-NEXT: return   $pop0{{$}}
16define i8* @copy_yes(i8* %dst, i8* %src, i32 %len) {
17  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
18  ret i8* %dst
19}
20
21; CHECK-LABEL: copy_no:
22; CHECK:      call     $drop=, memcpy, $0, $1, $2{{$}}
23; CHECK-NEXT: return{{$}}
24define void @copy_no(i8* %dst, i8* %src, i32 %len) {
25  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
26  ret void
27}
28
29; CHECK-LABEL: move_yes:
30; CHECK:      call     $push0=, memmove, $0, $1, $2{{$}}
31; CHECK-NEXT: return   $pop0{{$}}
32define i8* @move_yes(i8* %dst, i8* %src, i32 %len) {
33  call void @llvm.memmove.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
34  ret i8* %dst
35}
36
37; CHECK-LABEL: move_no:
38; CHECK:      call     $drop=, memmove, $0, $1, $2{{$}}
39; CHECK-NEXT: return{{$}}
40define void @move_no(i8* %dst, i8* %src, i32 %len) {
41  call void @llvm.memmove.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
42  ret void
43}
44
45; CHECK-LABEL: set_yes:
46; CHECK:      call     $push0=, memset, $0, $1, $2{{$}}
47; CHECK-NEXT: return   $pop0{{$}}
48define i8* @set_yes(i8* %dst, i8 %src, i32 %len) {
49  call void @llvm.memset.p0i8.i32(i8* %dst, i8 %src, i32 %len, i1 false)
50  ret i8* %dst
51}
52
53; CHECK-LABEL: set_no:
54; CHECK:      call     $drop=, memset, $0, $1, $2{{$}}
55; CHECK-NEXT: return{{$}}
56define void @set_no(i8* %dst, i8 %src, i32 %len) {
57  call void @llvm.memset.p0i8.i32(i8* %dst, i8 %src, i32 %len, i1 false)
58  ret void
59}
60
61
62; CHECK-LABEL: frame_index:
63; CHECK: call $drop=, memset, $pop{{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
64; CHECK: call $push{{[0-9]+}}=, memset, ${{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
65; CHECK: return{{$}}
66define void @frame_index() {
67entry:
68  %a = alloca [2048 x i8], align 16
69  %b = alloca [2048 x i8], align 16
70  %0 = getelementptr inbounds [2048 x i8], [2048 x i8]* %a, i32 0, i32 0
71  %1 = getelementptr inbounds [2048 x i8], [2048 x i8]* %b, i32 0, i32 0
72  call void @llvm.memset.p0i8.i32(i8* align 16 %0, i8 256, i32 1024, i1 false)
73  call void @llvm.memset.p0i8.i32(i8* align 16 %1, i8 256, i32 1024, i1 false)
74  ret void
75}
76
77; If the result value of memset doesn't get stackified, it should be marked
78; $drop. Note that we use a call to prevent tail dup so that we can test
79; this specific functionality.
80
81; CHECK-LABEL: drop_result:
82; CHECK: call $drop=, memset, $0, $1, $2
83declare i8* @def()
84declare void @block_tail_dup()
85define i8* @drop_result(i8* %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
86bb:
87  %tmp = icmp eq i32 %arg3, 0
88  br i1 %tmp, label %bb5, label %bb9
89
90bb5:
91  %tmp6 = icmp eq i32 %arg4, 0
92  br i1 %tmp6, label %bb7, label %bb8
93
94bb7:
95  call void @llvm.memset.p0i8.i32(i8* %arg, i8 %arg1, i32 %arg2, i1 false)
96  br label %bb11
97
98bb8:
99  br label %bb11
100
101bb9:
102  %tmp10 = call i8* @def()
103  br label %bb11
104
105bb11:
106  %tmp12 = phi i8* [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
107  call void @block_tail_dup()
108  ret i8* %tmp12
109}
110
111; This is the same as drop_result, except we let tail dup happen, so the
112; result of the memset *is* stackified.
113
114; CHECK-LABEL: tail_dup_to_reuse_result:
115; CHECK: call $push{{[0-9]+}}=, memset, $0, $1, $2
116define i8* @tail_dup_to_reuse_result(i8* %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
117bb:
118  %tmp = icmp eq i32 %arg3, 0
119  br i1 %tmp, label %bb5, label %bb9
120
121bb5:
122  %tmp6 = icmp eq i32 %arg4, 0
123  br i1 %tmp6, label %bb7, label %bb8
124
125bb7:
126  call void @llvm.memset.p0i8.i32(i8* %arg, i8 %arg1, i32 %arg2, i1 false)
127  br label %bb11
128
129bb8:
130  br label %bb11
131
132bb9:
133  %tmp10 = call i8* @def()
134  br label %bb11
135
136bb11:
137  %tmp12 = phi i8* [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
138  ret i8* %tmp12
139}
140