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
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=memcpyopt -S %s -verify-memoryssa | FileCheck %s
; Redundant memmove.
define i32 @redundant_memmove() {
; CHECK-LABEL: @redundant_memmove(
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT: [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT: ret i32 [[VAL]]
;
%array = alloca [26 x i32], align 16
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
%array.idx = getelementptr inbounds i8, ptr %array, i64 4
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
%val = load i32, ptr %array, align 16
ret i32 %val
}
; Used memmove, buffer is reset to zero.
define i32 @used_memmove_1() {
; CHECK-LABEL: @used_memmove_1(
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT: [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT: store i32 1, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT: call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 100, i1 false)
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT: ret i32 [[VAL]]
;
%array = alloca [26 x i32], align 16
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
%array.idx = getelementptr inbounds i8, ptr %array, i64 4
store i32 1, ptr %array.idx
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
%val = load i32, ptr %array.idx, align 4
ret i32 %val
}
; Used memmove.
define i32 @used_memmove_2() {
; CHECK-LABEL: @used_memmove_2(
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT: [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT: store i32 1, ptr [[ARRAY]], align 4
; CHECK-NEXT: call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 100, i1 false)
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT: ret i32 [[VAL]]
;
%array = alloca [26 x i32], align 16
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
%array.idx = getelementptr inbounds i8, ptr %array, i64 4
store i32 1, ptr %array
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
%val = load i32, ptr %array.idx, align 4
ret i32 %val
}
; Used memmove, buffer clobbered by opaque.
define i32 @used_memmove_3() {
; CHECK-LABEL: @used_memmove_3(
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [25 x i32], align 16
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 100, i1 false)
; CHECK-NEXT: call void @opaque(ptr [[ARRAY]])
; CHECK-NEXT: [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT: call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 96, i1 false)
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT: ret i32 [[VAL]]
;
%array = alloca [25 x i32], align 16
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 100, i1 false)
call void @opaque(ptr %array)
%array.idx = getelementptr inbounds i8, ptr %array, i64 4
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 96, i1 false)
%val = load i32, ptr %array, align 16
ret i32 %val
}
; Redundant memmove, not within the same basic block.
define i32 @redundant_memmove_different_bbs() {
; CHECK-LABEL: @redundant_memmove_different_bbs(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT: [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT: br label [[USE:%.*]]
; CHECK: use:
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT: ret i32 [[VAL]]
;
entry:
%array = alloca [26 x i32], align 16
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
%array.idx = getelementptr inbounds i8, ptr %array, i64 4
br label %use
use: ; preds = %entry
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
%val = load i32, ptr %array, align 16
ret i32 %val
}
@g_var = global [26 x i32] zeroinitializer, align 16
; Redundant memmove on a global variable.
define ptr @redundant_memmove_memset_global_variable() {
; CHECK-LABEL: @redundant_memmove_memset_global_variable(
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 @g_var, i8 0, i64 104, i1 false)
; CHECK-NEXT: ret ptr @g_var
;
call void @llvm.memset.p0.i64(ptr align 16 @g_var, i8 0, i64 104, i1 false)
call void @llvm.memmove.p0.p0.i64(ptr align 16 @g_var, ptr align 4 getelementptr inbounds nuw (i8, ptr @g_var, i64 4), i64 100, i1 false)
ret ptr @g_var
}
; Memset only partial.
define i32 @partial_memset() {
; CHECK-LABEL: @partial_memset(
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 92
; CHECK-NEXT: store i32 1, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 26, i1 false)
; CHECK-NEXT: [[ARRAY_IDX_2:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT: call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX_2]], i64 100, i1 false)
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT: ret i32 [[VAL]]
;
%array = alloca [26 x i32], align 16
%array.idx = getelementptr inbounds i8, ptr %array, i64 92
store i32 1, ptr %array.idx
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 26, i1 false)
%array.idx.2 = getelementptr inbounds i8, ptr %array, i64 4
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx.2, i64 100, i1 false)
%val = load i32, ptr %array.idx, align 4
ret i32 %val
}
; Memset length not constant.
define i32 @memset_length_not_constant(i64 %size) {
; CHECK-LABEL: @memset_length_not_constant(
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 [[SIZE:%.*]], i1 false)
; CHECK-NEXT: [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT: call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 100, i1 false)
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT: ret i32 [[VAL]]
;
%array = alloca [26 x i32], align 16
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 %size, i1 false)
%array.idx = getelementptr inbounds i8, ptr %array, i64 4
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
%val = load i32, ptr %array, align 16
ret i32 %val
}
; Memmove buffer not memset'd, different buffers.
define i32 @memset_memmove_dest_buffers_not_alias() {
; CHECK-LABEL: @memset_memmove_dest_buffers_not_alias(
; CHECK-NEXT: [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: [[ARRAY2:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT: [[ARRAY2_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY2]], i64 4
; CHECK-NEXT: call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY2]], ptr align 4 [[ARRAY2_IDX]], i64 100, i1 false)
; CHECK-NEXT: [[VAL:%.*]] = load i32, ptr [[ARRAY2]], align 16
; CHECK-NEXT: ret i32 [[VAL]]
;
%array = alloca [26 x i32], align 16
%array2 = alloca [26 x i32], align 16
call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
%array2.idx = getelementptr inbounds i8, ptr %array2, i64 4
call void @llvm.memmove.p0.p0.i64(ptr align 16 %array2, ptr align 4 %array2.idx, i64 100, i1 false)
%val = load i32, ptr %array2, align 16
ret i32 %val
}
declare void @opaque(ptr)
declare void @llvm.memset.p0.i64(ptr nocapture, i8, i64, i1)
declare void @llvm.memmove.p0.p0.i64(ptr nocapture, ptr nocapture, i64, i1)
|
