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
|
; RUN: opt < %s -S -indvars -loop-unroll -verify-loop-info | FileCheck %s
;
; Unit tests for loop unrolling using ScalarEvolution to compute trip counts.
;
; Indvars is run first to generate an "old" SCEV result. Some unit
; tests may check that SCEV is properly invalidated between passes.
; Completely unroll loops without a canonical IV.
;
; CHECK: @sansCanonical
; CHECK-NOT: phi
; CHECK-NOT: icmp
; CHECK: ret
define i32 @sansCanonical(i32* %base) nounwind {
entry:
br label %while.body
while.body:
%iv = phi i64 [ 10, %entry ], [ %iv.next, %while.body ]
%sum = phi i32 [ 0, %entry ], [ %sum.next, %while.body ]
%iv.next = add i64 %iv, -1
%adr = getelementptr inbounds i32* %base, i64 %iv.next
%tmp = load i32* %adr, align 8
%sum.next = add i32 %sum, %tmp
%iv.narrow = trunc i64 %iv.next to i32
%cmp.i65 = icmp sgt i32 %iv.narrow, 0
br i1 %cmp.i65, label %while.body, label %exit
exit:
ret i32 %sum
}
; SCEV unrolling properly handles loops with multiple exits. In this
; case, the computed trip count based on a canonical IV is *not* for a
; latch block. Canonical unrolling incorrectly unrolls it, but SCEV
; unrolling does not.
;
; CHECK: @earlyLoopTest
; CHECK: tail:
; CHECK-NOT: br
; CHECK: br i1 %cmp2, label %loop, label %exit2
define i64 @earlyLoopTest(i64* %base) nounwind {
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %inc, %tail ]
%s = phi i64 [ 0, %entry ], [ %s.next, %tail ]
%adr = getelementptr i64* %base, i64 %iv
%val = load i64* %adr
%s.next = add i64 %s, %val
%inc = add i64 %iv, 1
%cmp = icmp ne i64 %inc, 4
br i1 %cmp, label %tail, label %exit1
tail:
%cmp2 = icmp ne i64 %val, 0
br i1 %cmp2, label %loop, label %exit2
exit1:
ret i64 %s
exit2:
ret i64 %s.next
}
; SCEV properly unrolls multi-exit loops.
;
; CHECK: @multiExit
; CHECK: getelementptr i32* %base, i32 10
; CHECK-NEXT: load i32*
; CHECK: br i1 false, label %l2.10, label %exit1
; CHECK: l2.10:
; CHECK-NOT: br
; CHECK: ret i32
define i32 @multiExit(i32* %base) nounwind {
entry:
br label %l1
l1:
%iv1 = phi i32 [ 0, %entry ], [ %inc1, %l2 ]
%iv2 = phi i32 [ 0, %entry ], [ %inc2, %l2 ]
%inc1 = add i32 %iv1, 1
%inc2 = add i32 %iv2, 1
%adr = getelementptr i32* %base, i32 %iv1
%val = load i32* %adr
%cmp1 = icmp slt i32 %iv1, 5
br i1 %cmp1, label %l2, label %exit1
l2:
%cmp2 = icmp slt i32 %iv2, 10
br i1 %cmp2, label %l1, label %exit2
exit1:
ret i32 1
exit2:
ret i32 %val
}
; SCEV should not unroll a multi-exit loops unless the latch block has
; a known trip count, regardless of the early exit trip counts. The
; LoopUnroll utility uses this assumption to optimize the latch
; block's branch.
;
; CHECK: @multiExit
; CHECK: l3:
; CHECK-NOT: br
; CHECK: br i1 %cmp3, label %l1, label %exit3
define i32 @multiExitIncomplete(i32* %base) nounwind {
entry:
br label %l1
l1:
%iv1 = phi i32 [ 0, %entry ], [ %inc1, %l3 ]
%iv2 = phi i32 [ 0, %entry ], [ %inc2, %l3 ]
%inc1 = add i32 %iv1, 1
%inc2 = add i32 %iv2, 1
%adr = getelementptr i32* %base, i32 %iv1
%val = load i32* %adr
%cmp1 = icmp slt i32 %iv1, 5
br i1 %cmp1, label %l2, label %exit1
l2:
%cmp2 = icmp slt i32 %iv2, 10
br i1 %cmp2, label %l3, label %exit2
l3:
%cmp3 = icmp ne i32 %val, 0
br i1 %cmp3, label %l1, label %exit3
exit1:
ret i32 1
exit2:
ret i32 2
exit3:
ret i32 3
}
; When loop unroll merges a loop exit with one of its parent loop's
; exits, SCEV must forget its ExitNotTaken info.
;
; CHECK: @nestedUnroll
; CHECK-NOT: br i1
; CHECK: for.body87:
define void @nestedUnroll() nounwind {
entry:
br label %for.inc
for.inc:
br i1 false, label %for.inc, label %for.body38.preheader
for.body38.preheader:
br label %for.body38
for.body38:
%i.113 = phi i32 [ %inc76, %for.inc74 ], [ 0, %for.body38.preheader ]
%mul48 = mul nsw i32 %i.113, 6
br label %for.body43
for.body43:
%j.011 = phi i32 [ 0, %for.body38 ], [ %inc72, %for.body43 ]
%add49 = add nsw i32 %j.011, %mul48
%sh_prom50 = zext i32 %add49 to i64
%inc72 = add nsw i32 %j.011, 1
br i1 false, label %for.body43, label %for.inc74
for.inc74:
%inc76 = add nsw i32 %i.113, 1
br i1 false, label %for.body38, label %for.body87.preheader
for.body87.preheader:
br label %for.body87
for.body87:
br label %for.body87
}
|
