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 datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
6target triple = "wasm32-unknown-unknown"
7
8declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i1)
9declare void @llvm.memmove.p0i8.p0i8.i32(i8* nocapture, i8* nocapture readonly, i32, i1)
10declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i1)
11
12; Test that return values are optimized.
13
14; CHECK-LABEL: copy_yes:
15; CHECK:      call     $push0=, memcpy, $0, $1, $2{{$}}
16; CHECK-NEXT: return   $pop0{{$}}
17define i8* @copy_yes(i8* %dst, i8* %src, i32 %len) {
18  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
19  ret i8* %dst
20}
21
22; CHECK-LABEL: copy_no:
23; CHECK:      call     $drop=, memcpy, $0, $1, $2{{$}}
24; CHECK-NEXT: return{{$}}
25define void @copy_no(i8* %dst, i8* %src, i32 %len) {
26  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
27  ret void
28}
29
30; CHECK-LABEL: move_yes:
31; CHECK:      call     $push0=, memmove, $0, $1, $2{{$}}
32; CHECK-NEXT: return   $pop0{{$}}
33define i8* @move_yes(i8* %dst, i8* %src, i32 %len) {
34  call void @llvm.memmove.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
35  ret i8* %dst
36}
37
38; CHECK-LABEL: move_no:
39; CHECK:      call     $drop=, memmove, $0, $1, $2{{$}}
40; CHECK-NEXT: return{{$}}
41define void @move_no(i8* %dst, i8* %src, i32 %len) {
42  call void @llvm.memmove.p0i8.p0i8.i32(i8* %dst, i8* %src, i32 %len, i1 false)
43  ret void
44}
45
46; CHECK-LABEL: set_yes:
47; CHECK:      call     $push0=, memset, $0, $1, $2{{$}}
48; CHECK-NEXT: return   $pop0{{$}}
49define i8* @set_yes(i8* %dst, i8 %src, i32 %len) {
50  call void @llvm.memset.p0i8.i32(i8* %dst, i8 %src, i32 %len, i1 false)
51  ret i8* %dst
52}
53
54; CHECK-LABEL: set_no:
55; CHECK:      call     $drop=, memset, $0, $1, $2{{$}}
56; CHECK-NEXT: return{{$}}
57define void @set_no(i8* %dst, i8 %src, i32 %len) {
58  call void @llvm.memset.p0i8.i32(i8* %dst, i8 %src, i32 %len, i1 false)
59  ret void
60}
61
62
63; CHECK-LABEL: frame_index:
64; CHECK: call $drop=, memset, $pop{{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
65; CHECK: call $push{{[0-9]+}}=, memset, ${{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
66; CHECK: return{{$}}
67define void @frame_index() {
68entry:
69  %a = alloca [2048 x i8], align 16
70  %b = alloca [2048 x i8], align 16
71  %0 = getelementptr inbounds [2048 x i8], [2048 x i8]* %a, i32 0, i32 0
72  %1 = getelementptr inbounds [2048 x i8], [2048 x i8]* %b, i32 0, i32 0
73  call void @llvm.memset.p0i8.i32(i8* align 16 %0, i8 256, i32 1024, i1 false)
74  call void @llvm.memset.p0i8.i32(i8* align 16 %1, i8 256, i32 1024, i1 false)
75  ret void
76}
77
78; If the result value of memset doesn't get stackified, it should be marked
79; $drop. Note that we use a call to prevent tail dup so that we can test
80; this specific functionality.
81
82; CHECK-LABEL: drop_result:
83; CHECK: call $drop=, memset, $0, $1, $2
84declare i8* @def()
85declare void @block_tail_dup()
86define i8* @drop_result(i8* %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
87bb:
88  %tmp = icmp eq i32 %arg3, 0
89  br i1 %tmp, label %bb5, label %bb9
90
91bb5:
92  %tmp6 = icmp eq i32 %arg4, 0
93  br i1 %tmp6, label %bb7, label %bb8
94
95bb7:
96  call void @llvm.memset.p0i8.i32(i8* %arg, i8 %arg1, i32 %arg2, i1 false)
97  br label %bb11
98
99bb8:
100  br label %bb11
101
102bb9:
103  %tmp10 = call i8* @def()
104  br label %bb11
105
106bb11:
107  %tmp12 = phi i8* [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
108  call void @block_tail_dup()
109  ret i8* %tmp12
110}
111
112; This is the same as drop_result, except we let tail dup happen, so the
113; result of the memset *is* stackified.
114
115; CHECK-LABEL: tail_dup_to_reuse_result:
116; CHECK: call $push{{[0-9]+}}=, memset, $0, $1, $2
117define i8* @tail_dup_to_reuse_result(i8* %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
118bb:
119  %tmp = icmp eq i32 %arg3, 0
120  br i1 %tmp, label %bb5, label %bb9
121
122bb5:
123  %tmp6 = icmp eq i32 %arg4, 0
124  br i1 %tmp6, label %bb7, label %bb8
125
126bb7:
127  call void @llvm.memset.p0i8.i32(i8* %arg, i8 %arg1, i32 %arg2, i1 false)
128  br label %bb11
129
130bb8:
131  br label %bb11
132
133bb9:
134  %tmp10 = call i8* @def()
135  br label %bb11
136
137bb11:
138  %tmp12 = phi i8* [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
139  ret i8* %tmp12
140}
141