File: mem-intrinsics.ll

package info (click to toggle)
llvm-toolchain-16 1%3A16.0.6-15~deb12u1
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,634,792 kB
  • sloc: cpp: 6,179,261; ansic: 1,216,205; asm: 741,319; python: 196,614; objc: 75,325; f90: 49,640; lisp: 32,396; pascal: 12,286; sh: 9,394; perl: 7,442; ml: 5,494; awk: 3,523; makefile: 2,723; javascript: 1,206; xml: 886; fortran: 581; cs: 573
file content (137 lines) | stat: -rw-r--r-- 4,071 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
; RUN: llc < %s -asm-verbose=false -mcpu=mvp -disable-wasm-fallthrough-return-opt -wasm-disable-explicit-locals -wasm-keep-registers -tail-dup-placement=0 | FileCheck %s

; Test memcpy, memmove, and memset intrinsics.

target triple = "wasm32-unknown-unknown"

declare void @llvm.memcpy.p0.p0.i32(ptr nocapture, ptr nocapture readonly, i32, i1)
declare void @llvm.memmove.p0.p0.i32(ptr nocapture, ptr nocapture readonly, i32, i1)
declare void @llvm.memset.p0.i32(ptr nocapture, i8, i32, i1)

; Test that return values are optimized.

; CHECK-LABEL: copy_yes:
; CHECK:      call     $push0=, memcpy, $0, $1, $2{{$}}
; CHECK-NEXT: return   $pop0{{$}}
define ptr @copy_yes(ptr %dst, ptr %src, i32 %len) {
  call void @llvm.memcpy.p0.p0.i32(ptr %dst, ptr %src, i32 %len, i1 false)
  ret ptr %dst
}

; CHECK-LABEL: copy_no:
; CHECK:      call     $drop=, memcpy, $0, $1, $2{{$}}
; CHECK-NEXT: return{{$}}
define void @copy_no(ptr %dst, ptr %src, i32 %len) {
  call void @llvm.memcpy.p0.p0.i32(ptr %dst, ptr %src, i32 %len, i1 false)
  ret void
}

; CHECK-LABEL: move_yes:
; CHECK:      call     $push0=, memmove, $0, $1, $2{{$}}
; CHECK-NEXT: return   $pop0{{$}}
define ptr @move_yes(ptr %dst, ptr %src, i32 %len) {
  call void @llvm.memmove.p0.p0.i32(ptr %dst, ptr %src, i32 %len, i1 false)
  ret ptr %dst
}

; CHECK-LABEL: move_no:
; CHECK:      call     $drop=, memmove, $0, $1, $2{{$}}
; CHECK-NEXT: return{{$}}
define void @move_no(ptr %dst, ptr %src, i32 %len) {
  call void @llvm.memmove.p0.p0.i32(ptr %dst, ptr %src, i32 %len, i1 false)
  ret void
}

; CHECK-LABEL: set_yes:
; CHECK:      call     $push0=, memset, $0, $1, $2{{$}}
; CHECK-NEXT: return   $pop0{{$}}
define ptr @set_yes(ptr %dst, i8 %src, i32 %len) {
  call void @llvm.memset.p0.i32(ptr %dst, i8 %src, i32 %len, i1 false)
  ret ptr %dst
}

; CHECK-LABEL: set_no:
; CHECK:      call     $drop=, memset, $0, $1, $2{{$}}
; CHECK-NEXT: return{{$}}
define void @set_no(ptr %dst, i8 %src, i32 %len) {
  call void @llvm.memset.p0.i32(ptr %dst, i8 %src, i32 %len, i1 false)
  ret void
}


; CHECK-LABEL: frame_index:
; CHECK: call $drop=, memset, $pop{{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
; CHECK: call $push{{[0-9]+}}=, memset, ${{[0-9]+}}, $pop{{[0-9]+}}, $pop{{[0-9]+}}{{$}}
; CHECK: return{{$}}
define void @frame_index() {
entry:
  %a = alloca [2048 x i8], align 16
  %b = alloca [2048 x i8], align 16
  call void @llvm.memset.p0.i32(ptr align 16 %a, i8 256, i32 1024, i1 false)
  call void @llvm.memset.p0.i32(ptr align 16 %b, i8 256, i32 1024, i1 false)
  ret void
}

; If the result value of memset doesn't get stackified, it should be marked
; $drop. Note that we use a call to prevent tail dup so that we can test
; this specific functionality.

; CHECK-LABEL: drop_result:
; CHECK: call $drop=, memset, $0, $1, $2
declare ptr @def()
declare void @block_tail_dup()
define ptr @drop_result(ptr %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
bb:
  %tmp = icmp eq i32 %arg3, 0
  br i1 %tmp, label %bb5, label %bb9

bb5:
  %tmp6 = icmp eq i32 %arg4, 0
  br i1 %tmp6, label %bb7, label %bb8

bb7:
  call void @llvm.memset.p0.i32(ptr %arg, i8 %arg1, i32 %arg2, i1 false)
  br label %bb11

bb8:
  br label %bb11

bb9:
  %tmp10 = call ptr @def()
  br label %bb11

bb11:
  %tmp12 = phi ptr [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
  call void @block_tail_dup()
  ret ptr %tmp12
}

; This is the same as drop_result, except we let tail dup happen, so the
; result of the memset *is* stackified.

; CHECK-LABEL: tail_dup_to_reuse_result:
; CHECK: call $push{{[0-9]+}}=, memset, $0, $1, $2
define ptr @tail_dup_to_reuse_result(ptr %arg, i8 %arg1, i32 %arg2, i32 %arg3, i32 %arg4) {
bb:
  %tmp = icmp eq i32 %arg3, 0
  br i1 %tmp, label %bb5, label %bb9

bb5:
  %tmp6 = icmp eq i32 %arg4, 0
  br i1 %tmp6, label %bb7, label %bb8

bb7:
  call void @llvm.memset.p0.i32(ptr %arg, i8 %arg1, i32 %arg2, i1 false)
  br label %bb11

bb8:
  br label %bb11

bb9:
  %tmp10 = call ptr @def()
  br label %bb11

bb11:
  %tmp12 = phi ptr [ %arg, %bb7 ], [ %arg, %bb8 ], [ %tmp10, %bb9 ]
  ret ptr %tmp12
}