1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt < %s -basic-aa -slp-vectorizer -dce -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
4 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128"
5 target triple = "i386-apple-macosx10.9.0"
7 ; We disable the vectorization of <3 x float> for now
9 ; float foo(float *A) {
14 ; for (int i=0; i < 121; i+=3) {
23 define float @foo(float* nocapture readonly %A) {
26 ; CHECK-NEXT: [[TMP0:%.*]] = bitcast float* [[A:%.*]] to <2 x float>*
27 ; CHECK-NEXT: [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 4
28 ; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 2
29 ; CHECK-NEXT: [[TMP2:%.*]] = load float, float* [[ARRAYIDX2]], align 4
30 ; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x float> [[TMP1]], i32 0
31 ; CHECK-NEXT: br label [[FOR_BODY:%.*]]
33 ; CHECK-NEXT: [[TMP4:%.*]] = phi float [ [[TMP3]], [[ENTRY:%.*]] ], [ [[DOTPRE:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE:%.*]] ]
34 ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ]
35 ; CHECK-NEXT: [[B_032:%.*]] = phi float [ [[TMP2]], [[ENTRY]] ], [ [[ADD14:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ]
36 ; CHECK-NEXT: [[TMP5:%.*]] = phi <2 x float> [ [[TMP1]], [[ENTRY]] ], [ [[TMP11:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ]
37 ; CHECK-NEXT: [[TMP6:%.*]] = add nsw i64 [[INDVARS_IV]], 1
38 ; CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP6]]
39 ; CHECK-NEXT: [[TMP7:%.*]] = load float, float* [[ARRAYIDX7]], align 4
40 ; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x float> poison, float [[TMP4]], i32 0
41 ; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x float> [[TMP8]], float [[TMP7]], i32 1
42 ; CHECK-NEXT: [[TMP10:%.*]] = fmul <2 x float> [[TMP9]], <float 7.000000e+00, float 8.000000e+00>
43 ; CHECK-NEXT: [[TMP11]] = fadd <2 x float> [[TMP5]], [[TMP10]]
44 ; CHECK-NEXT: [[TMP12:%.*]] = add nsw i64 [[INDVARS_IV]], 2
45 ; CHECK-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]]
46 ; CHECK-NEXT: [[TMP13:%.*]] = load float, float* [[ARRAYIDX12]], align 4
47 ; CHECK-NEXT: [[MUL13:%.*]] = fmul float [[TMP13]], 9.000000e+00
48 ; CHECK-NEXT: [[ADD14]] = fadd float [[B_032]], [[MUL13]]
49 ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 3
50 ; CHECK-NEXT: [[TMP14:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
51 ; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[TMP14]], 121
52 ; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY_FOR_BODY_CRIT_EDGE]], label [[FOR_END:%.*]]
53 ; CHECK: for.body.for.body_crit_edge:
54 ; CHECK-NEXT: [[ARRAYIDX3_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV_NEXT]]
55 ; CHECK-NEXT: [[DOTPRE]] = load float, float* [[ARRAYIDX3_PHI_TRANS_INSERT]], align 4
56 ; CHECK-NEXT: br label [[FOR_BODY]]
58 ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <2 x float> [[TMP11]], i32 0
59 ; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x float> [[TMP11]], i32 1
60 ; CHECK-NEXT: [[ADD16:%.*]] = fadd float [[TMP15]], [[TMP16]]
61 ; CHECK-NEXT: [[ADD17:%.*]] = fadd float [[ADD16]], [[ADD14]]
62 ; CHECK-NEXT: ret float [[ADD17]]
65 %0 = load float, float* %A, align 4
66 %arrayidx1 = getelementptr inbounds float, float* %A, i64 1
67 %1 = load float, float* %arrayidx1, align 4
68 %arrayidx2 = getelementptr inbounds float, float* %A, i64 2
69 %2 = load float, float* %arrayidx2, align 4
72 for.body: ; preds = %for.body.for.body_crit_edge, %entry
73 %3 = phi float [ %0, %entry ], [ %.pre, %for.body.for.body_crit_edge ]
74 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body.for.body_crit_edge ]
75 %B.032 = phi float [ %2, %entry ], [ %add14, %for.body.for.body_crit_edge ]
76 %G.031 = phi float [ %1, %entry ], [ %add9, %for.body.for.body_crit_edge ]
77 %R.030 = phi float [ %0, %entry ], [ %add4, %for.body.for.body_crit_edge ]
78 %mul = fmul float %3, 7.000000e+00
79 %add4 = fadd float %R.030, %mul
80 %4 = add nsw i64 %indvars.iv, 1
81 %arrayidx7 = getelementptr inbounds float, float* %A, i64 %4
82 %5 = load float, float* %arrayidx7, align 4
83 %mul8 = fmul float %5, 8.000000e+00
84 %add9 = fadd float %G.031, %mul8
85 %6 = add nsw i64 %indvars.iv, 2
86 %arrayidx12 = getelementptr inbounds float, float* %A, i64 %6
87 %7 = load float, float* %arrayidx12, align 4
88 %mul13 = fmul float %7, 9.000000e+00
89 %add14 = fadd float %B.032, %mul13
90 %indvars.iv.next = add i64 %indvars.iv, 3
91 %8 = trunc i64 %indvars.iv.next to i32
92 %cmp = icmp slt i32 %8, 121
93 br i1 %cmp, label %for.body.for.body_crit_edge, label %for.end
95 for.body.for.body_crit_edge: ; preds = %for.body
96 %arrayidx3.phi.trans.insert = getelementptr inbounds float, float* %A, i64 %indvars.iv.next
97 %.pre = load float, float* %arrayidx3.phi.trans.insert, align 4
100 for.end: ; preds = %for.body
101 %add16 = fadd float %add4, %add9
102 %add17 = fadd float %add16, %add14