1 ; RUN: opt -loop-vectorize -mtriple=x86_64-apple-macosx -S -mcpu=corei7-avx -enable-interleaved-mem-accesses=false < %s | FileCheck %s
2 target 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"
4 @kernel = global [512 x float] zeroinitializer, align 16
5 @kernel2 = global [512 x float] zeroinitializer, align 16
6 @kernel3 = global [512 x float] zeroinitializer, align 16
7 @kernel4 = global [512 x float] zeroinitializer, align 16
8 @src_data = global [1536 x float] zeroinitializer, align 16
9 @r_ = global i8 0, align 1
10 @g_ = global i8 0, align 1
11 @b_ = global i8 0, align 1
13 ; We don't want to vectorize most loops containing gathers because they are
14 ; expensive. This function represents a point where vectorization starts to
16 ; Make sure we are conservative and don't vectorize it.
19 define void @_Z4testmm(i64 %size, i64 %offset) {
21 %cmp53 = icmp eq i64 %size, 0
22 br i1 %cmp53, label %for.end, label %for.body.lr.ph
28 %r.057 = phi float [ 0.000000e+00, %for.body.lr.ph ], [ %add10, %for.body ]
29 %g.056 = phi float [ 0.000000e+00, %for.body.lr.ph ], [ %add20, %for.body ]
30 %v.055 = phi i64 [ 0, %for.body.lr.ph ], [ %inc, %for.body ]
31 %b.054 = phi float [ 0.000000e+00, %for.body.lr.ph ], [ %add30, %for.body ]
32 %add = add i64 %v.055, %offset
33 %mul = mul i64 %add, 3
34 %arrayidx = getelementptr inbounds [1536 x float], [1536 x float]* @src_data, i64 0, i64 %mul
35 %0 = load float, float* %arrayidx, align 4
36 %arrayidx2 = getelementptr inbounds [512 x float], [512 x float]* @kernel, i64 0, i64 %v.055
37 %1 = load float, float* %arrayidx2, align 4
38 %mul3 = fmul fast float %0, %1
39 %arrayidx4 = getelementptr inbounds [512 x float], [512 x float]* @kernel2, i64 0, i64 %v.055
40 %2 = load float, float* %arrayidx4, align 4
41 %mul5 = fmul fast float %mul3, %2
42 %arrayidx6 = getelementptr inbounds [512 x float], [512 x float]* @kernel3, i64 0, i64 %v.055
43 %3 = load float, float* %arrayidx6, align 4
44 %mul7 = fmul fast float %mul5, %3
45 %arrayidx8 = getelementptr inbounds [512 x float], [512 x float]* @kernel4, i64 0, i64 %v.055
46 %4 = load float, float* %arrayidx8, align 4
47 %mul9 = fmul fast float %mul7, %4
48 %add10 = fadd fast float %r.057, %mul9
49 %arrayidx.sum = add i64 %mul, 1
50 %arrayidx11 = getelementptr inbounds [1536 x float], [1536 x float]* @src_data, i64 0, i64 %arrayidx.sum
51 %5 = load float, float* %arrayidx11, align 4
52 %mul13 = fmul fast float %1, %5
53 %mul15 = fmul fast float %2, %mul13
54 %mul17 = fmul fast float %3, %mul15
55 %mul19 = fmul fast float %4, %mul17
56 %add20 = fadd fast float %g.056, %mul19
57 %arrayidx.sum52 = add i64 %mul, 2
58 %arrayidx21 = getelementptr inbounds [1536 x float], [1536 x float]* @src_data, i64 0, i64 %arrayidx.sum52
59 %6 = load float, float* %arrayidx21, align 4
60 %mul23 = fmul fast float %1, %6
61 %mul25 = fmul fast float %2, %mul23
62 %mul27 = fmul fast float %3, %mul25
63 %mul29 = fmul fast float %4, %mul27
64 %add30 = fadd fast float %b.054, %mul29
65 %inc = add i64 %v.055, 1
66 %exitcond = icmp ne i64 %inc, %size
67 br i1 %exitcond, label %for.body, label %for.cond.for.end_crit_edge
69 for.cond.for.end_crit_edge:
70 %add30.lcssa = phi float [ %add30, %for.body ]
71 %add20.lcssa = phi float [ %add20, %for.body ]
72 %add10.lcssa = phi float [ %add10, %for.body ]
73 %phitmp = fptoui float %add10.lcssa to i8
74 %phitmp60 = fptoui float %add20.lcssa to i8
75 %phitmp61 = fptoui float %add30.lcssa to i8
79 %r.0.lcssa = phi i8 [ %phitmp, %for.cond.for.end_crit_edge ], [ 0, %entry ]
80 %g.0.lcssa = phi i8 [ %phitmp60, %for.cond.for.end_crit_edge ], [ 0, %entry ]
81 %b.0.lcssa = phi i8 [ %phitmp61, %for.cond.for.end_crit_edge ], [ 0, %entry ]
82 store i8 %r.0.lcssa, i8* @r_, align 1
83 store i8 %g.0.lcssa, i8* @g_, align 1
84 store i8 %b.0.lcssa, i8* @b_, align 1