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
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -indvars -dce -S | FileCheck %s
; Provide legal integer types.
target datalayout = "n8:16:32:64"
@A = external global i32
;; Convert a pre-increment check on the latch into a post increment check
define i32 @pre_to_post_add() {
; CHECK-LABEL: @pre_to_post_add(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[I_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; CHECK-NEXT: store i32 [[I]], i32* @A
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[I_NEXT]], 1001
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[LOOPEXIT:%.*]]
; CHECK: loopexit:
; CHECK-NEXT: ret i32 1000
;
entry:
br label %loop
loop:
%i = phi i32 [ 0, %entry ], [ %i.next, %loop ]
%i.next = add i32 %i, 1
store i32 %i, i32* @A
%c = icmp slt i32 %i, 1000
br i1 %c, label %loop, label %loopexit
loopexit:
ret i32 %i
}
; TODO: we should be able to convert the subtract into a post-decrement check
define i32 @pre_to_post_sub() {
; CHECK-LABEL: @pre_to_post_sub(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ 1000, [[ENTRY:%.*]] ], [ [[I_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[I_NEXT]] = sub nsw i32 [[I]], 1
; CHECK-NEXT: store i32 [[I]], i32* @A
; CHECK-NEXT: [[C:%.*]] = icmp ugt i32 [[I]], 0
; CHECK-NEXT: br i1 [[C]], label [[LOOP]], label [[LOOPEXIT:%.*]]
; CHECK: loopexit:
; CHECK-NEXT: ret i32 0
;
entry:
br label %loop
loop:
%i = phi i32 [ 1000, %entry ], [ %i.next, %loop ]
%i.next = sub i32 %i, 1
store i32 %i, i32* @A
%c = icmp sgt i32 %i, 0
br i1 %c, label %loop, label %loopexit
loopexit:
ret i32 %i
}
; LFTR should eliminate the need for the computation of i*i completely. It
; is only used to compute the exit value.
define i32 @quadratic_slt() {
; CHECK-LABEL: @quadratic_slt(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ 7, [[ENTRY:%.*]] ], [ [[I_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; CHECK-NEXT: store i32 [[I]], i32* @A
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[I_NEXT]], 33
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[LOOPEXIT:%.*]]
; CHECK: loopexit:
; CHECK-NEXT: ret i32 32
;
entry:
br label %loop
loop:
%i = phi i32 [ 7, %entry ], [ %i.next, %loop ]
%i.next = add i32 %i, 1
store i32 %i, i32* @A
%i2 = mul i32 %i, %i
%c = icmp slt i32 %i2, 1000
br i1 %c, label %loop, label %loopexit
loopexit:
ret i32 %i
}
; Same as previous but with sle test
define i32 @quadratic_sle() {
; CHECK-LABEL: @quadratic_sle(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ 7, [[ENTRY:%.*]] ], [ [[I_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; CHECK-NEXT: store i32 [[I]], i32* @A
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[I_NEXT]], 33
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[LOOPEXIT:%.*]]
; CHECK: loopexit:
; CHECK-NEXT: ret i32 32
;
entry:
br label %loop
loop:
%i = phi i32 [ 7, %entry ], [ %i.next, %loop ]
%i.next = add i32 %i, 1
store i32 %i, i32* @A
%i2 = mul i32 %i, %i
%c = icmp sle i32 %i2, 1000
br i1 %c, label %loop, label %loopexit
loopexit:
ret i32 %i
}
; Same as previous but with ule test
define i32 @quadratic_ule() {
; CHECK-LABEL: @quadratic_ule(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ 7, [[ENTRY:%.*]] ], [ [[I_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; CHECK-NEXT: store i32 [[I]], i32* @A
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[I_NEXT]], 33
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[LOOPEXIT:%.*]]
; CHECK: loopexit:
; CHECK-NEXT: ret i32 32
;
entry:
br label %loop
loop:
%i = phi i32 [ 7, %entry ], [ %i.next, %loop ]
%i.next = add i32 %i, 1
store i32 %i, i32* @A
%i2 = mul i32 %i, %i
%c = icmp ule i32 %i2, 1000
br i1 %c, label %loop, label %loopexit
loopexit:
ret i32 %i
}
@data = common global [240 x i8] zeroinitializer, align 16
define void @test_zext(i8* %a) #0 {
; CHECK-LABEL: @test_zext(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[P_0:%.*]] = phi i8* [ getelementptr inbounds ([240 x i8], [240 x i8]* @data, i64 0, i64 0), [[ENTRY:%.*]] ], [ [[TMP3:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[DOT0:%.*]] = phi i8* [ [[A:%.*]], [[ENTRY]] ], [ [[TMP:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[TMP]] = getelementptr inbounds i8, i8* [[DOT0]], i64 1
; CHECK-NEXT: [[TMP2:%.*]] = load i8, i8* [[DOT0]], align 1
; CHECK-NEXT: [[TMP3]] = getelementptr inbounds i8, i8* [[P_0]], i64 1
; CHECK-NEXT: store i8 [[TMP2]], i8* [[P_0]], align 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i8* [[TMP3]], getelementptr (i8, i8* getelementptr inbounds ([240 x i8], [240 x i8]* @data, i64 0, i64 0), i64 240)
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%i.0 = phi i8 [ 0, %entry ], [ %tmp4, %loop ]
%p.0 = phi i8* [ getelementptr inbounds ([240 x i8], [240 x i8]* @data, i64 0, i64 0), %entry ], [ %tmp3, %loop ]
%.0 = phi i8* [ %a, %entry ], [ %tmp, %loop ]
%tmp = getelementptr inbounds i8, i8* %.0, i64 1
%tmp2 = load i8, i8* %.0, align 1
%tmp3 = getelementptr inbounds i8, i8* %p.0, i64 1
store i8 %tmp2, i8* %p.0, align 1
%tmp4 = add i8 %i.0, 1
%tmp5 = icmp ult i8 %tmp4, -16
br i1 %tmp5, label %loop, label %exit
exit:
ret void
}
; It is okay to do LFTR on this loop even though the trip count is a
; division because in this case the division can be optimized to a
; shift.
define void @test_udiv_as_shift(i8* %a, i8 %n) nounwind uwtable ssp {
; CHECK-LABEL: @test_udiv_as_shift(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[E:%.*]] = icmp sgt i8 [[N:%.*]], 3
; CHECK-NEXT: br i1 [[E]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = add i8 [[N]], 3
; CHECK-NEXT: [[TMP1:%.*]] = lshr i8 [[TMP0]], 2
; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[TMP1]], 1
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I1:%.*]] = phi i8 [ [[I1_INC:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
; CHECK-NEXT: [[I1_INC]] = add nuw nsw i8 [[I1]], 1
; CHECK-NEXT: store volatile i8 0, i8* [[A:%.*]]
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i8 [[I1_INC]], [[TMP2]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
%e = icmp sgt i8 %n, 3
br i1 %e, label %loop, label %exit
loop:
%i = phi i8 [ 0, %entry ], [ %i.inc, %loop ]
%i1 = phi i8 [ 0, %entry ], [ %i1.inc, %loop ]
%i.inc = add nsw i8 %i, 4
%i1.inc = add i8 %i1, 1
store volatile i8 0, i8* %a
%c = icmp slt i8 %i, %n
br i1 %c, label %loop, label %exit
exit:
ret void
}
; Don't RAUW the loop's original comparison instruction if it has other uses
; which aren't dominated by the new comparison instruction (which we insert
; at the branch user).
define void @use_before_branch() {
; CHECK-LABEL: @use_before_branch(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOPENTRY_0:%.*]]
; CHECK: loopentry.0:
; CHECK-NEXT: [[MB_Y_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[TMP_152:%.*]], [[LOOPENTRY_1:%.*]] ]
; CHECK-NEXT: [[TMP_14:%.*]] = icmp ule i32 [[MB_Y_0]], 3
; CHECK-NEXT: br i1 [[TMP_14]], label [[LOOPENTRY_1]], label [[LOOPEXIT_0:%.*]]
; CHECK: loopentry.1:
; CHECK-NEXT: [[TMP_152]] = add nuw nsw i32 [[MB_Y_0]], 2
; CHECK-NEXT: br label [[LOOPENTRY_0]]
; CHECK: loopexit.0:
; CHECK-NEXT: unreachable
;
entry:
br label %loopentry.0
loopentry.0:
%mb_y.0 = phi i32 [ 0, %entry ], [ %tmp.152, %loopentry.1 ]
%tmp.14 = icmp sle i32 %mb_y.0, 3
%tmp.15 = zext i1 %tmp.14 to i32
br i1 %tmp.14, label %loopentry.1, label %loopexit.0
loopentry.1:
%tmp.152 = add i32 %mb_y.0, 2
br label %loopentry.0
loopexit.0: ; preds = %loopentry.0
unreachable
}
@.str3 = private constant [6 x i8] c"%lld\0A\00", align 1
declare i32 @printf(i8* noalias nocapture, ...) nounwind
; PR13371: indvars pass incorrectly substitutes 'undef' values
;
; LFTR should not user %undef as the loop counter.
define i64 @no_undef_counter() nounwind {
; CHECK-LABEL: @no_undef_counter(
; CHECK-NEXT: func_start:
; CHECK-NEXT: br label [[BLOCK9:%.*]]
; CHECK: block9:
; CHECK-NEXT: [[UNDEF:%.*]] = phi i64 [ [[NEXT_UNDEF:%.*]], [[BLOCK9]] ], [ undef, [[FUNC_START:%.*]] ]
; CHECK-NEXT: [[ITER:%.*]] = phi i64 [ [[NEXT_ITER:%.*]], [[BLOCK9]] ], [ 1, [[FUNC_START]] ]
; CHECK-NEXT: [[NEXT_ITER]] = add nuw nsw i64 [[ITER]], 1
; CHECK-NEXT: [[TMP0:%.*]] = tail call i32 (i8*, ...) @printf(i8* noalias nocapture getelementptr inbounds ([6 x i8], [6 x i8]* @.str3, i64 0, i64 0), i64 [[NEXT_ITER]], i64 [[UNDEF]])
; CHECK-NEXT: [[NEXT_UNDEF]] = add nsw i64 [[UNDEF]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[NEXT_ITER]], 100
; CHECK-NEXT: br i1 [[EXITCOND]], label [[BLOCK9]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret i64 0
;
func_start:
br label %block9
block9: ; preds = %block9,%func_start
%undef = phi i64 [ %next_undef, %block9 ], [ undef, %func_start ]
%iter = phi i64 [ %next_iter, %block9 ], [ 1, %func_start ]
%next_iter = add nsw i64 %iter, 1
%0 = tail call i32 (i8*, ...) @printf(i8* noalias nocapture getelementptr inbounds ([6 x i8], [6 x i8]* @.str3, i64 0, i64 0), i64 %next_iter, i64 %undef)
%next_undef = add nsw i64 %undef, 1
%_tmp_3 = icmp slt i64 %next_iter, 100
br i1 %_tmp_3, label %block9, label %exit
exit: ; preds = %block9
ret i64 0
}
|
