1; RUN: opt -basicaa -loop-vectorize -S -mcpu=core-avx2 < %s | FileCheck %s
2
3;float inc = 0.5;
4;void foo(float *A, unsigned N) {
5;
6;  for (unsigned i=0; i<N; i++){
7;    A[i] += inc;
8;  }
9;}
10
11; CHECK-LABEL: foo
12; CHECK: vector.body
13; CHECK: load <8 x float>
14; CHECK: fadd <8 x float>
15; CHECK: store <8 x float>
16
17target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
18target triple = "x86_64-unknown-linux-gnu"
19
20@inc = global float 5.000000e-01, align 4
21
22define void @foo(float* nocapture %A, i32 %N) #0 {
23entry:
24  %cmp3 = icmp eq i32 %N, 0
25  br i1 %cmp3, label %for.end, label %for.body.preheader
26
27for.body.preheader:                               ; preds = %entry
28  br label %for.body
29
30for.body:                                         ; preds = %for.body.preheader, %for.body
31  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
32  %0 = load float, float* @inc, align 4
33  %arrayidx = getelementptr inbounds float, float* %A, i64 %indvars.iv
34  %1 = load float, float* %arrayidx, align 4
35  %add = fadd float %0, %1
36  store float %add, float* %arrayidx, align 4
37  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
38  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
39  %exitcond = icmp eq i32 %lftr.wideiv, %N
40  br i1 %exitcond, label %for.end.loopexit, label %for.body
41
42for.end.loopexit:                                 ; preds = %for.body
43  br label %for.end
44
45for.end:                                          ; preds = %for.end.loopexit, %entry
46  ret void
47}
48