1; RUN: opt < %s -debug-only=loop-vectorize -loop-vectorize -vectorizer-maximize-bandwidth -O2 -S 2>&1 | FileCheck %s
2; REQUIRES: asserts
3
4target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
5target triple = "x86_64-unknown-linux-gnu"
6
7@a = global [1024 x i8] zeroinitializer, align 16
8@b = global [1024 x i8] zeroinitializer, align 16
9
10define i32 @foo() {
11; This function has a loop of SAD pattern. Here we check when VF = 16 the
12; register usage doesn't exceed 16.
13;
14; CHECK-LABEL: foo
15; CHECK:      LV(REG): VF = 4
16; CHECK-NEXT: LV(REG): Found max usage: 4
17; CHECK:      LV(REG): VF = 8
18; CHECK-NEXT: LV(REG): Found max usage: 7
19; CHECK:      LV(REG): VF = 16
20; CHECK-NEXT: LV(REG): Found max usage: 13
21
22entry:
23  br label %for.body
24
25for.cond.cleanup:
26  %add.lcssa = phi i32 [ %add, %for.body ]
27  ret i32 %add.lcssa
28
29for.body:
30  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
31  %s.015 = phi i32 [ 0, %entry ], [ %add, %for.body ]
32  %arrayidx = getelementptr inbounds [1024 x i8], [1024 x i8]* @a, i64 0, i64 %indvars.iv
33  %0 = load i8, i8* %arrayidx, align 1
34  %conv = zext i8 %0 to i32
35  %arrayidx2 = getelementptr inbounds [1024 x i8], [1024 x i8]* @b, i64 0, i64 %indvars.iv
36  %1 = load i8, i8* %arrayidx2, align 1
37  %conv3 = zext i8 %1 to i32
38  %sub = sub nsw i32 %conv, %conv3
39  %ispos = icmp sgt i32 %sub, -1
40  %neg = sub nsw i32 0, %sub
41  %2 = select i1 %ispos, i32 %sub, i32 %neg
42  %add = add nsw i32 %2, %s.015
43  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
44  %exitcond = icmp eq i64 %indvars.iv.next, 1024
45  br i1 %exitcond, label %for.cond.cleanup, label %for.body
46}
47
48define i64 @bar(i64* nocapture %a) {
49; CHECK-LABEL: bar
50; CHECK:       LV(REG): VF = 2
51; CHECK:       LV(REG): Found max usage: 4
52;
53entry:
54  br label %for.body
55
56for.cond.cleanup:
57  %add2.lcssa = phi i64 [ %add2, %for.body ]
58  ret i64 %add2.lcssa
59
60for.body:
61  %i.012 = phi i64 [ 0, %entry ], [ %inc, %for.body ]
62  %s.011 = phi i64 [ 0, %entry ], [ %add2, %for.body ]
63  %arrayidx = getelementptr inbounds i64, i64* %a, i64 %i.012
64  %0 = load i64, i64* %arrayidx, align 8
65  %add = add nsw i64 %0, %i.012
66  store i64 %add, i64* %arrayidx, align 8
67  %add2 = add nsw i64 %add, %s.011
68  %inc = add nuw nsw i64 %i.012, 1
69  %exitcond = icmp eq i64 %inc, 1024
70  br i1 %exitcond, label %for.cond.cleanup, label %for.body
71}
72