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
|
; RUN: opt %loadPolly -polly-optree -analyze < %s | FileCheck %s -match-full-lines
;
; Do not try to forward %i.trunc, it is not synthesizable in %body.
;
define void @func(i32 %n, i32* noalias nonnull %A) {
entry:
br label %for
for:
%j = phi i32 [0, %entry], [%j.inc, %inc]
%j.cmp = icmp slt i32 %j, %n
%zero = sext i32 0 to i64
br i1 %j.cmp, label %inner.for, label %exit
; This loop has some unusual properties:
; * It has a known iteration count (8), therefore SCoP-compatible.
; * %i.trunc is synthesizable within the loop ({1,+,1}<%while.body>).
; * %i.trunc is not synthesizable outside of the loop, because its value is
; unknown when exiting.
; (should be 8, but ScalarEvolution currently seems unable to derive that)
;
; ScalarEvolution currently seems to not able to handle the %zero.
; If it becomes more intelligent, there might be other such loop constructs.
inner.for:
%i = phi i64 [%zero, %for], [%i.inc, %inner.for]
%i.inc = add nuw nsw i64 %i, 1
%i.trunc = trunc i64 %i.inc to i32
%i.and = and i32 %i.trunc, 7
%inner.cond = icmp eq i32 %i.and, 0
br i1 %inner.cond, label %body, label %inner.for
body:
store i32 %i.trunc, i32* %A
br label %inc
inc:
%j.inc = add nuw nsw i32 %j, 1
br label %for
exit:
br label %return
return:
ret void
}
; CHECK: ForwardOpTree executed, but did not modify anything
|
