1; RUN: opt < %s -loop-vectorize -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -debug-only=loop-vectorize -stats -S -vectorizer-min-trip-count=21 2>&1 | FileCheck %s
2; REQUIRES: asserts
3
4; CHECK: LV: Loop hints: force=enabled
5; CHECK: LV: Loop hints: force=?
6; No more loops in the module
7; CHECK-NOT: LV: Loop hints: force=
8; CHECK: 2 loop-vectorize               - Number of loops analyzed for vectorization
9; CHECK: 1 loop-vectorize               - Number of loops vectorized
10
11target 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"
12target triple = "x86_64-apple-macosx10.8.0"
13
14;
15; The source code for the test:
16;
17; void foo(float* restrict A, float* restrict B)
18; {
19;     for (int i = 0; i < 20; ++i) A[i] += B[i];
20; }
21;
22
23;
24; This loop will be vectorized, although the trip count is below the threshold, but vectorization is explicitly forced in metadata.
25;
26define void @vectorized(float* noalias nocapture %A, float* noalias nocapture readonly %B) {
27entry:
28  br label %for.body
29
30for.body:
31  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
32  %arrayidx = getelementptr inbounds float, float* %B, i64 %indvars.iv
33  %0 = load float, float* %arrayidx, align 4, !llvm.mem.parallel_loop_access !1
34  %arrayidx2 = getelementptr inbounds float, float* %A, i64 %indvars.iv
35  %1 = load float, float* %arrayidx2, align 4, !llvm.mem.parallel_loop_access !1
36  %add = fadd fast float %0, %1
37  store float %add, float* %arrayidx2, align 4, !llvm.mem.parallel_loop_access !1
38  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
39  %exitcond = icmp eq i64 %indvars.iv.next, 20
40  br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !1
41
42for.end:
43  ret void
44}
45
46!1 = !{!1, !2}
47!2 = !{!"llvm.loop.vectorize.enable", i1 true}
48
49;
50; This loop will not be vectorized as the trip count is below the threshold.
51;
52define void @not_vectorized(float* noalias nocapture %A, float* noalias nocapture readonly %B) {
53entry:
54  br label %for.body
55
56for.body:
57  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
58  %arrayidx = getelementptr inbounds float, float* %B, i64 %indvars.iv
59  %0 = load float, float* %arrayidx, align 4, !llvm.mem.parallel_loop_access !3
60  %arrayidx2 = getelementptr inbounds float, float* %A, i64 %indvars.iv
61  %1 = load float, float* %arrayidx2, align 4, !llvm.mem.parallel_loop_access !3
62  %add = fadd fast float %0, %1
63  store float %add, float* %arrayidx2, align 4, !llvm.mem.parallel_loop_access !3
64  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
65  %exitcond = icmp eq i64 %indvars.iv.next, 20
66  br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !3
67
68for.end:
69  ret void
70}
71
72!3 = !{!3}
73
74