• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: opt < %s -loop-vectorize -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-vector-width=4 -S | FileCheck %s
2
3; Check that a counting-down loop which has no primary induction variable
4; is vectorized with preferred predication.
5
6; CHECK-LABEL: vector.body:
7; CHECK-LABEL: middle.block:
8; CHECK-NEXT:    br i1 true,
9
10target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
11
12define dso_local void @foo(i8* noalias nocapture readonly %A, i8* noalias nocapture readonly %B, i8* noalias nocapture %C, i32 %N) {
13entry:
14  %cmp6 = icmp eq i32 %N, 0
15  br i1 %cmp6, label %while.end, label %while.body.preheader
16
17while.body.preheader:
18  br label %while.body
19
20while.body:
21  %N.addr.010 = phi i32 [ %dec, %while.body ], [ %N, %while.body.preheader ]
22  %C.addr.09 = phi i8* [ %incdec.ptr4, %while.body ], [ %C, %while.body.preheader ]
23  %B.addr.08 = phi i8* [ %incdec.ptr1, %while.body ], [ %B, %while.body.preheader ]
24  %A.addr.07 = phi i8* [ %incdec.ptr, %while.body ], [ %A, %while.body.preheader ]
25  %incdec.ptr = getelementptr inbounds i8, i8* %A.addr.07, i32 1
26  %0 = load i8, i8* %A.addr.07, align 1
27  %incdec.ptr1 = getelementptr inbounds i8, i8* %B.addr.08, i32 1
28  %1 = load i8, i8* %B.addr.08, align 1
29  %add = add i8 %1, %0
30  %incdec.ptr4 = getelementptr inbounds i8, i8* %C.addr.09, i32 1
31  store i8 %add, i8* %C.addr.09, align 1
32  %dec = add i32 %N.addr.010, -1
33  %cmp = icmp eq i32 %dec, 0
34  br i1 %cmp, label %while.end.loopexit, label %while.body
35
36while.end.loopexit:
37  br label %while.end
38
39while.end:
40  ret void
41}
42
43; Make sure a loop is successfully vectorized with fold-tail when the backedge
44; taken count is constant and used inside the loop. Issue revealed by D76992.
45;
46define void @reuse_const_btc(i8* %A) optsize {
47; CHECK-LABEL: @reuse_const_btc
48; CHECK: {{%.*}} = icmp ule <4 x i32> {{%.*}}, <i32 13, i32 13, i32 13, i32 13>
49; CHECK: {{%.*}} = select <4 x i1> {{%.*}}, <4 x i32> <i32 12, i32 12, i32 12, i32 12>, <4 x i32> <i32 13, i32 13, i32 13, i32 13>
50;
51entry:
52  br label %loop
53
54loop:
55  %riv = phi i32 [ 13, %entry ], [ %rivMinus1, %merge ]
56  %sub = sub nuw nsw i32 20, %riv
57  %arrayidx = getelementptr inbounds i8, i8* %A, i32 %sub
58  %cond0 = icmp eq i32 %riv, 7
59  br i1 %cond0, label %then, label %else
60then:
61  br label %merge
62else:
63  br label %merge
64merge:
65  %blend = phi i32 [ 13, %then ], [ 12, %else ]
66  %trunc = trunc i32 %blend to i8
67  store i8 %trunc, i8* %arrayidx, align 1
68  %rivMinus1 = add nuw nsw i32 %riv, -1
69  %cond = icmp eq i32 %riv, 0
70  br i1 %cond, label %exit, label %loop
71
72exit:
73  ret void
74}
75