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
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
|
; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s
; REQUIRES: asserts
target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
target triple = "thumbv8.1m.main-arm-none-eabi"
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %arrayidx = getelementptr inbounds i16, ptr %s, i32 %i.016
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %1 = load i16, ptr %arrayidx, align 2
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv = sext i16 %1 to i32
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br i1 %cmp2, label %if.then, label %for.inc
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %conv6 = add i16 %1, %0
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %arrayidx7 = getelementptr inbounds i16, ptr %d, i32 %i.016
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %conv6, ptr %arrayidx7, align 2
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br label %for.inc
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %inc = add nuw nsw i32 %i.016, 1
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %exitcond.not = icmp eq i32 %inc, %n
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
; CHECK: LV: Scalar loop costs: 5.
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %arrayidx = getelementptr inbounds i16, ptr %s, i32 %i.016
; CHECK: LV: Found an estimated cost of 18 for VF 2 For instruction: %1 = load i16, ptr %arrayidx, align 2
; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction: %conv = sext i16 %1 to i32
; CHECK: LV: Found an estimated cost of 20 for VF 2 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br i1 %cmp2, label %if.then, label %for.inc
; CHECK: LV: Found an estimated cost of 26 for VF 2 For instruction: %conv6 = add i16 %1, %0
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %arrayidx7 = getelementptr inbounds i16, ptr %d, i32 %i.016
; CHECK: LV: Found an estimated cost of 16 for VF 2 For instruction: store i16 %conv6, ptr %arrayidx7, align 2
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br label %for.inc
; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %inc = add nuw nsw i32 %i.016, 1
; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %exitcond.not = icmp eq i32 %inc, %n
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
; CHECK: LV: Vector loop of width 2 costs: 43.
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %arrayidx = getelementptr inbounds i16, ptr %s, i32 %i.016
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %1 = load i16, ptr %arrayidx, align 2
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv = sext i16 %1 to i32
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br i1 %cmp2, label %if.then, label %for.inc
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %conv6 = add i16 %1, %0
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %arrayidx7 = getelementptr inbounds i16, ptr %d, i32 %i.016
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: store i16 %conv6, ptr %arrayidx7, align 2
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br label %for.inc
; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %inc = add nuw nsw i32 %i.016, 1
; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %exitcond.not = icmp eq i32 %inc, %n
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
; CHECK: LV: Vector loop of width 4 costs: 2.
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %arrayidx = getelementptr inbounds i16, ptr %s, i32 %i.016
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %1 = load i16, ptr %arrayidx, align 2
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv = sext i16 %1 to i32
; CHECK: LV: Found an estimated cost of 36 for VF 8 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br i1 %cmp2, label %if.then, label %for.inc
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv6 = add i16 %1, %0
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %arrayidx7 = getelementptr inbounds i16, ptr %d, i32 %i.016
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: store i16 %conv6, ptr %arrayidx7, align 2
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br label %for.inc
; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %inc = add nuw nsw i32 %i.016, 1
; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %exitcond.not = icmp eq i32 %inc, %n
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
; CHECK: LV: Vector loop of width 8 costs: 5.
; CHECK: LV: Selecting VF: 4.
define void @expensive_icmp(ptr noalias nocapture %d, ptr nocapture readonly %s, i32 %n, i16 zeroext %m) #0 {
entry:
%cmp15 = icmp sgt i32 %n, 0
br i1 %cmp15, label %for.body.lr.ph, label %for.cond.cleanup
for.body.lr.ph: ; preds = %entry
%conv1 = zext i16 %m to i32
%0 = trunc i32 %n to i16
br label %for.body
for.cond.cleanup: ; preds = %for.inc, %entry
ret void
for.body: ; preds = %for.body.lr.ph, %for.inc
%i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
%arrayidx = getelementptr inbounds i16, ptr %s, i32 %i.016
%1 = load i16, ptr %arrayidx, align 2
%conv = sext i16 %1 to i32
%cmp2 = icmp sgt i32 %conv, %conv1
br i1 %cmp2, label %if.then, label %for.inc
if.then: ; preds = %for.body
%conv6 = add i16 %1, %0
%arrayidx7 = getelementptr inbounds i16, ptr %d, i32 %i.016
store i16 %conv6, ptr %arrayidx7, align 2
br label %for.inc
for.inc: ; preds = %for.body, %if.then
%inc = add nuw nsw i32 %i.016, 1
%exitcond.not = icmp eq i32 %inc, %n
br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
}
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %pSrcA.addr.011 = phi ptr [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %incdec.ptr = getelementptr inbounds i8, ptr %pSrcA.addr.011, i32 1
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %0 = load i8, ptr %pSrcA.addr.011, align 1
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv1 = sext i8 %0 to i32
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %1 = load i8, ptr %pSrcB.addr.09, align 1
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv3 = sext i8 %1 to i32
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %mul = mul nsw i32 %conv3, %conv1
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %shr = ashr i32 %mul, 7
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %2 = icmp slt i32 %shr, 127
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv4 = trunc i32 %spec.select.i to i8
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %incdec.ptr5 = getelementptr inbounds i8, ptr %pDst.addr.010, i32 1
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %conv4, ptr %pDst.addr.010, align 1
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %dec = add i32 %blkCnt.012, -1
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %cmp.not = icmp eq i32 %dec, 0
; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
; CHECK: LV: Scalar loop costs: 9.
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %pSrcA.addr.011 = phi ptr [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %incdec.ptr = getelementptr inbounds i8, ptr %pSrcA.addr.011, i32 1
; CHECK: LV: Found an estimated cost of 18 for VF 2 For instruction: %0 = load i8, ptr %pSrcA.addr.011, align 1
; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction: %conv1 = sext i8 %0 to i32
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: LV: Found an estimated cost of 18 for VF 2 For instruction: %1 = load i8, ptr %pSrcB.addr.09, align 1
; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction: %conv3 = sext i8 %1 to i32
; CHECK: LV: Found an estimated cost of 26 for VF 2 For instruction: %mul = mul nsw i32 %conv3, %conv1
; CHECK: LV: Found an estimated cost of 18 for VF 2 For instruction: %shr = ashr i32 %mul, 7
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %2 = icmp slt i32 %shr, 127
; CHECK: LV: Found an estimated cost of 22 for VF 2 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %conv4 = trunc i32 %spec.select.i to i8
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %incdec.ptr5 = getelementptr inbounds i8, ptr %pDst.addr.010, i32 1
; CHECK: LV: Found an estimated cost of 18 for VF 2 For instruction: store i8 %conv4, ptr %pDst.addr.010, align 1
; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %dec = add i32 %blkCnt.012, -1
; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %cmp.not = icmp eq i32 %dec, 0
; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
; CHECK: LV: Vector loop of width 2 costs: 65.
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %pSrcA.addr.011 = phi ptr [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %incdec.ptr = getelementptr inbounds i8, ptr %pSrcA.addr.011, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %0 = load i8, ptr %pSrcA.addr.011, align 1
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv1 = sext i8 %0 to i32
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %1 = load i8, ptr %pSrcB.addr.09, align 1
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv3 = sext i8 %1 to i32
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %mul = mul nsw i32 %conv3, %conv1
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %shr = ashr i32 %mul, 7
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %2 = icmp slt i32 %shr, 127
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv4 = trunc i32 %spec.select.i to i8
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %incdec.ptr5 = getelementptr inbounds i8, ptr %pDst.addr.010, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: store i8 %conv4, ptr %pDst.addr.010, align 1
; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %dec = add i32 %blkCnt.012, -1
; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %cmp.not = icmp eq i32 %dec, 0
; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
; CHECK: LV: Vector loop of width 4 costs: 3.
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %pSrcA.addr.011 = phi ptr [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %incdec.ptr = getelementptr inbounds i8, ptr %pSrcA.addr.011, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %0 = load i8, ptr %pSrcA.addr.011, align 1
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv1 = sext i8 %0 to i32
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %1 = load i8, ptr %pSrcB.addr.09, align 1
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv3 = sext i8 %1 to i32
; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %mul = mul nsw i32 %conv3, %conv1
; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %shr = ashr i32 %mul, 7
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %2 = icmp slt i32 %shr, 127
; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv4 = trunc i32 %spec.select.i to i8
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %incdec.ptr5 = getelementptr inbounds i8, ptr %pDst.addr.010, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: store i8 %conv4, ptr %pDst.addr.010, align 1
; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %dec = add i32 %blkCnt.012, -1
; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %cmp.not = icmp eq i32 %dec, 0
; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
; CHECK: LV: Vector loop of width 8 costs: 3.
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %pSrcA.addr.011 = phi ptr [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %incdec.ptr = getelementptr inbounds i8, ptr %pSrcA.addr.011, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction: %0 = load i8, ptr %pSrcA.addr.011, align 1
; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction: %conv1 = sext i8 %0 to i32
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction: %1 = load i8, ptr %pSrcB.addr.09, align 1
; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction: %conv3 = sext i8 %1 to i32
; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction: %mul = mul nsw i32 %conv3, %conv1
; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction: %shr = ashr i32 %mul, 7
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %2 = icmp slt i32 %shr, 127
; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction: %conv4 = trunc i32 %spec.select.i to i8
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %incdec.ptr5 = getelementptr inbounds i8, ptr %pDst.addr.010, i32 1
; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction: store i8 %conv4, ptr %pDst.addr.010, align 1
; CHECK: LV: Found an estimated cost of 1 for VF 16 For instruction: %dec = add i32 %blkCnt.012, -1
; CHECK: LV: Found an estimated cost of 1 for VF 16 For instruction: %cmp.not = icmp eq i32 %dec, 0
; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
; CHECK: LV: Vector loop of width 16 costs: 3.
; CHECK: LV: Selecting VF: 16.
define void @cheap_icmp(ptr nocapture readonly %pSrcA, ptr nocapture readonly %pSrcB, ptr nocapture %pDst, i32 %blockSize) #0 {
entry:
%cmp.not8 = icmp eq i32 %blockSize, 0
br i1 %cmp.not8, label %while.end, label %while.body.preheader
while.body.preheader: ; preds = %entry
br label %while.body
while.body: ; preds = %while.body.preheader, %while.body
%blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
%pSrcA.addr.011 = phi ptr [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
%pDst.addr.010 = phi ptr [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
%pSrcB.addr.09 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
%incdec.ptr = getelementptr inbounds i8, ptr %pSrcA.addr.011, i32 1
%0 = load i8, ptr %pSrcA.addr.011, align 1
%conv1 = sext i8 %0 to i32
%incdec.ptr2 = getelementptr inbounds i8, ptr %pSrcB.addr.09, i32 1
%1 = load i8, ptr %pSrcB.addr.09, align 1
%conv3 = sext i8 %1 to i32
%mul = mul nsw i32 %conv3, %conv1
%shr = ashr i32 %mul, 7
%2 = icmp slt i32 %shr, 127
%spec.select.i = select i1 %2, i32 %shr, i32 127
%conv4 = trunc i32 %spec.select.i to i8
%incdec.ptr5 = getelementptr inbounds i8, ptr %pDst.addr.010, i32 1
store i8 %conv4, ptr %pDst.addr.010, align 1
%dec = add i32 %blkCnt.012, -1
%cmp.not = icmp eq i32 %dec, 0
br i1 %cmp.not, label %while.end.loopexit, label %while.body
while.end.loopexit: ; preds = %while.body
br label %while.end
while.end: ; preds = %while.end.loopexit, %entry
ret void
}
; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %cmp1 = fcmp
; CHECK: LV: Found an estimated cost of 12 for VF 2 For instruction: %cmp1 = fcmp
; CHECK: LV: Found an estimated cost of 24 for VF 4 For instruction: %cmp1 = fcmp
define void @floatcmp(ptr nocapture readonly %pSrc, ptr nocapture %pDst, i32 %blockSize) #0 {
entry:
%cmp.not7 = icmp eq i32 %blockSize, 0
br i1 %cmp.not7, label %while.end, label %while.body
while.body: ; preds = %entry, %while.body
%pSrc.addr.010 = phi ptr [ %incdec.ptr2, %while.body ], [ %pSrc, %entry ]
%blockSize.addr.09 = phi i32 [ %dec, %while.body ], [ %blockSize, %entry ]
%pDst.addr.08 = phi ptr [ %incdec.ptr, %while.body ], [ %pDst, %entry ]
%0 = load float, ptr %pSrc.addr.010, align 4
%cmp1 = fcmp nnan ninf nsz olt float %0, 0.000000e+00
%cond = select nnan ninf nsz i1 %cmp1, float 1.000000e+01, float %0
%conv = fptosi float %cond to i32
%incdec.ptr = getelementptr inbounds i32, ptr %pDst.addr.08, i32 1
store i32 %conv, ptr %pDst.addr.08, align 4
%incdec.ptr2 = getelementptr inbounds float, ptr %pSrc.addr.010, i32 1
%dec = add i32 %blockSize.addr.09, -1
%cmp.not = icmp eq i32 %dec, 0
br i1 %cmp.not, label %while.end, label %while.body
while.end: ; preds = %while.body, %entry
ret void
}
attributes #0 = { "target-features"="+mve" }
|
