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
181
182
183
184
| ; RUN: llc < %s -march=nvptx64 -mcpu=sm_35 -O0 | FileCheck %s --check-prefix PTX
; RUN: opt < %s -S -nvptx-lower-aggr-copies | FileCheck %s --check-prefix IR
; Verify that the NVPTXLowerAggrCopies pass works as expected - calls to
; llvm.mem* intrinsics get lowered to loops.
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "nvptx64-unknown-unknown"
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i1) #1
declare void @llvm.memmove.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i1) #1
declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i1) #1
define i8* @memcpy_caller(i8* %dst, i8* %src, i64 %n) #0 {
entry:
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %n, i1 false)
ret i8* %dst
; IR-LABEL: @memcpy_caller
; IR: entry:
; IR: [[Cond:%[0-9]+]] = icmp ne i64 %n, 0
; IR: br i1 [[Cond]], label %loop-memcpy-expansion, label %post-loop-memcpy-expansion
; IR: loop-memcpy-expansion:
; IR: %loop-index = phi i64 [ 0, %entry ], [ [[IndexInc:%[0-9]+]], %loop-memcpy-expansion ]
; IR: [[SrcGep:%[0-9]+]] = getelementptr inbounds i8, i8* %src, i64 %loop-index
; IR: [[Load:%[0-9]+]] = load i8, i8* [[SrcGep]]
; IR: [[DstGep:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64 %loop-index
; IR: store i8 [[Load]], i8* [[DstGep]]
; IR: [[IndexInc]] = add i64 %loop-index, 1
; IR: [[Cond2:%[0-9]+]] = icmp ult i64 [[IndexInc]], %n
; IR: br i1 [[Cond2]], label %loop-memcpy-expansion, label %post-loop-memcpy-expansion
; IR-LABEL: post-loop-memcpy-expansion:
; IR: ret i8* %dst
; PTX-LABEL: .visible .func (.param .b64 func_retval0) memcpy_caller
; PTX: LBB[[LABEL:[_0-9]+]]:
; PTX: ld.u8 %rs[[REG:[0-9]+]]
; PTX: st.u8 [%rd{{[0-9]+}}], %rs[[REG]]
; PTX: add.s64 %rd[[COUNTER:[0-9]+]], %rd{{[0-9]+}}, 1
; PTX: setp.lt.u64 %p[[PRED:[0-9]+]], %rd[[COUNTER]], %rd
; PTX: @%p[[PRED]] bra LBB[[LABEL]]
}
define i8* @memcpy_volatile_caller(i8* %dst, i8* %src, i64 %n) #0 {
entry:
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %n, i1 true)
ret i8* %dst
; IR-LABEL: @memcpy_volatile_caller
; IR: entry:
; IR: [[Cond:%[0-9]+]] = icmp ne i64 %n, 0
; IR: br i1 [[Cond]], label %loop-memcpy-expansion, label %post-loop-memcpy-expansion
; IR: loop-memcpy-expansion:
; IR: %loop-index = phi i64 [ 0, %entry ], [ [[IndexInc:%[0-9]+]], %loop-memcpy-expansion ]
; IR: [[SrcGep:%[0-9]+]] = getelementptr inbounds i8, i8* %src, i64 %loop-index
; IR: [[Load:%[0-9]+]] = load volatile i8, i8* [[SrcGep]]
; IR: [[DstGep:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64 %loop-index
; IR: store volatile i8 [[Load]], i8* [[DstGep]]
; IR: [[IndexInc]] = add i64 %loop-index, 1
; IR: [[Cond2:%[0-9]+]] = icmp ult i64 [[IndexInc]], %n
; IR: br i1 [[Cond2]], label %loop-memcpy-expansion, label %post-loop-memcpy-expansion
; IR-LABEL: post-loop-memcpy-expansion:
; IR: ret i8* %dst
; PTX-LABEL: .visible .func (.param .b64 func_retval0) memcpy_volatile_caller
; PTX: LBB[[LABEL:[_0-9]+]]:
; PTX: ld.volatile.u8 %rs[[REG:[0-9]+]]
; PTX: st.volatile.u8 [%rd{{[0-9]+}}], %rs[[REG]]
; PTX: add.s64 %rd[[COUNTER:[0-9]+]], %rd{{[0-9]+}}, 1
; PTX: setp.lt.u64 %p[[PRED:[0-9]+]], %rd[[COUNTER]], %rd
; PTX: @%p[[PRED]] bra LBB[[LABEL]]
}
define i8* @memcpy_casting_caller(i32* %dst, i32* %src, i64 %n) #0 {
entry:
%0 = bitcast i32* %dst to i8*
%1 = bitcast i32* %src to i8*
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %0, i8* %1, i64 %n, i1 false)
ret i8* %0
; Check that casts in calls to memcpy are handled properly
; IR-LABEL: @memcpy_casting_caller
; IR: [[DSTCAST:%[0-9]+]] = bitcast i32* %dst to i8*
; IR: [[SRCCAST:%[0-9]+]] = bitcast i32* %src to i8*
; IR: getelementptr inbounds i8, i8* [[SRCCAST]]
; IR: getelementptr inbounds i8, i8* [[DSTCAST]]
}
define i8* @memcpy_known_size(i8* %dst, i8* %src) {
entry:
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 144, i1 false)
ret i8* %dst
; Check that calls with compile-time constant size are handled correctly
; IR-LABEL: @memcpy_known_size
; IR: entry:
; IR: br label %load-store-loop
; IR: load-store-loop:
; IR: %loop-index = phi i64 [ 0, %entry ], [ [[IndexInc:%[0-9]+]], %load-store-loop ]
; IR: [[SrcGep:%[0-9]+]] = getelementptr inbounds i8, i8* %src, i64 %loop-index
; IR: [[Load:%[0-9]+]] = load i8, i8* [[SrcGep]]
; IR: [[DstGep:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64 %loop-index
; IR: store i8 [[Load]], i8* [[DstGep]]
; IR: [[IndexInc]] = add i64 %loop-index, 1
; IR: [[Cond:%[0-9]+]] = icmp ult i64 %3, 144
; IR: br i1 [[Cond]], label %load-store-loop, label %memcpy-split
}
define i8* @memset_caller(i8* %dst, i32 %c, i64 %n) #0 {
entry:
%0 = trunc i32 %c to i8
tail call void @llvm.memset.p0i8.i64(i8* %dst, i8 %0, i64 %n, i1 false)
ret i8* %dst
; IR-LABEL: @memset_caller
; IR: [[VAL:%[0-9]+]] = trunc i32 %c to i8
; IR: [[CMPREG:%[0-9]+]] = icmp eq i64 0, %n
; IR: br i1 [[CMPREG]], label %split, label %loadstoreloop
; IR: loadstoreloop:
; IR: [[STOREPTR:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64
; IR-NEXT: store i8 [[VAL]], i8* [[STOREPTR]]
; PTX-LABEL: .visible .func (.param .b64 func_retval0) memset_caller(
; PTX: ld.param.u32 %r[[C:[0-9]+]]
; PTX: cvt.u16.u32 %rs[[REG:[0-9]+]], %r[[C]];
; PTX: LBB[[LABEL:[_0-9]+]]:
; PTX: st.u8 [%rd{{[0-9]+}}], %rs[[REG]]
; PTX: add.s64 %rd[[COUNTER:[0-9]+]], %rd{{[0-9]+}}, 1
; PTX: setp.lt.u64 %p[[PRED:[0-9]+]], %rd[[COUNTER]], %rd
; PTX: @%p[[PRED]] bra LBB[[LABEL]]
}
define i8* @volatile_memset_caller(i8* %dst, i32 %c, i64 %n) #0 {
entry:
%0 = trunc i32 %c to i8
tail call void @llvm.memset.p0i8.i64(i8* %dst, i8 %0, i64 %n, i1 true)
ret i8* %dst
; IR-LABEL: @volatile_memset_caller
; IR: [[VAL:%[0-9]+]] = trunc i32 %c to i8
; IR: loadstoreloop:
; IR: [[STOREPTR:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64
; IR-NEXT: store volatile i8 [[VAL]], i8* [[STOREPTR]]
}
define i8* @memmove_caller(i8* %dst, i8* %src, i64 %n) #0 {
entry:
tail call void @llvm.memmove.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %n, i1 false)
ret i8* %dst
; IR-LABEL: @memmove_caller
; IR: icmp ult i8* %src, %dst
; IR: [[PHIVAL:%[0-9a-zA-Z_]+]] = phi i64
; IR-NEXT: %index_ptr = sub i64 [[PHIVAL]], 1
; IR: [[FWDPHIVAL:%[0-9a-zA-Z_]+]] = phi i64
; IR: {{%[0-9a-zA-Z_]+}} = add i64 [[FWDPHIVAL]], 1
; PTX-LABEL: .visible .func (.param .b64 func_retval0) memmove_caller(
; PTX: ld.param.u64 %rd[[N:[0-9]+]]
; PTX-DAG: setp.eq.s64 %p[[NEQ0:[0-9]+]], %rd[[N]], 0
; PTX-DAG: setp.ge.u64 %p[[SRC_GT_THAN_DST:[0-9]+]], %rd{{[0-9]+}}, %rd{{[0-9]+}}
; PTX-NEXT: @%p[[SRC_GT_THAN_DST]] bra LBB[[FORWARD_BB:[0-9_]+]]
; -- this is the backwards copying BB
; PTX: @%p[[NEQ0]] bra LBB[[EXIT:[0-9_]+]]
; PTX: add.s64 %rd{{[0-9]}}, %rd{{[0-9]}}, -1
; PTX: ld.u8 %rs[[ELEMENT:[0-9]+]]
; PTX: st.u8 [%rd{{[0-9]+}}], %rs[[ELEMENT]]
; -- this is the forwards copying BB
; PTX: LBB[[FORWARD_BB]]:
; PTX: @%p[[NEQ0]] bra LBB[[EXIT]]
; PTX: ld.u8 %rs[[ELEMENT2:[0-9]+]]
; PTX: st.u8 [%rd{{[0-9]+}}], %rs[[ELEMENT2]]
; PTX: add.s64 %rd{{[0-9]+}}, %rd{{[0-9]+}}, 1
; -- exit block
; PTX: LBB[[EXIT]]:
; PTX-NEXT: st.param.b64 [func_retval0
; PTX-NEXT: ret
}
|