1; RUN: opt < %s  -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -dce -instcombine -S | FileCheck %s
2
3target 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"
4target triple = "x86_64-apple-macosx10.8.0"
5
6;CHECK-LABEL: @flags1(
7;CHECK: load <4 x i32>
8;CHECK: mul nsw <4 x i32>
9;CHECK: store <4 x i32>
10;CHECK: ret i32
11define i32 @flags1(i32 %n, i32* nocapture %A) nounwind uwtable ssp {
12  %1 = icmp sgt i32 %n, 9
13  br i1 %1, label %.lr.ph, label %._crit_edge
14
15.lr.ph:                                           ; preds = %0, %.lr.ph
16  %indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 9, %0 ]
17  %2 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv
18  %3 = load i32, i32* %2, align 4
19  %4 = mul nsw i32 %3, 3
20  store i32 %4, i32* %2, align 4
21  %indvars.iv.next = add i64 %indvars.iv, 1
22  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
23  %exitcond = icmp eq i32 %lftr.wideiv, %n
24  br i1 %exitcond, label %._crit_edge, label %.lr.ph
25
26._crit_edge:                                      ; preds = %.lr.ph, %0
27  ret i32 undef
28}
29
30
31;CHECK-LABEL: @flags2(
32;CHECK: load <4 x i32>
33;CHECK: mul <4 x i32>
34;CHECK: store <4 x i32>
35;CHECK: ret i32
36define i32 @flags2(i32 %n, i32* nocapture %A) nounwind uwtable ssp {
37  %1 = icmp sgt i32 %n, 9
38  br i1 %1, label %.lr.ph, label %._crit_edge
39
40.lr.ph:                                           ; preds = %0, %.lr.ph
41  %indvars.iv = phi i64 [ %indvars.iv.next, %.lr.ph ], [ 9, %0 ]
42  %2 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv
43  %3 = load i32, i32* %2, align 4
44  %4 = mul i32 %3, 3
45  store i32 %4, i32* %2, align 4
46  %indvars.iv.next = add i64 %indvars.iv, 1
47  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
48  %exitcond = icmp eq i32 %lftr.wideiv, %n
49  br i1 %exitcond, label %._crit_edge, label %.lr.ph
50
51._crit_edge:                                      ; preds = %.lr.ph, %0
52  ret i32 undef
53}
54
55; Make sure we copy fast math flags and use them for the final reduction.
56; CHECK-LABEL: fast_math
57; CHECK: load <4 x float>
58; CHECK: fadd fast <4 x float>
59; CHECK: br
60; CHECK: fadd fast <4 x float>
61; CHECK: fadd fast <4 x float>
62define float @fast_math(float* noalias %s) {
63entry:
64  br label %for.body
65
66for.body:
67  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
68  %q.04 = phi float [ 0.000000e+00, %entry ], [ %add, %for.body ]
69  %arrayidx = getelementptr inbounds float, float* %s, i64 %indvars.iv
70  %0 = load float, float* %arrayidx, align 4
71  %add = fadd fast float %q.04, %0
72  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
73  %exitcond = icmp eq i64 %indvars.iv.next, 256
74  br i1 %exitcond, label %for.end, label %for.body
75
76for.end:
77  %add.lcssa = phi float [ %add, %for.body ]
78  ret float %add.lcssa
79}
80