• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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
3
4target 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"
5target triple = "i386-apple-macosx10.9.0"
6
7; We disable the vectorization of <3 x float> for now
8
9; float foo(float *A) {
10;
11;   float R = A[0];
12;   float G = A[1];
13;   float B = A[2];
14;   for (int i=0; i < 121; i+=3) {
15;     R+=A[i+0]*7;
16;     G+=A[i+1]*8;
17;     B+=A[i+2]*9;
18;   }
19;
20;   return R+G+B;
21; }
22
23define float @foo(float* nocapture readonly %A) {
24; CHECK-LABEL: @foo(
25; CHECK-NEXT:  entry:
26; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[A:%.*]], align 4
27; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds float, float* [[A]], i64 1
28; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[ARRAYIDX1]], align 4
29; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 2
30; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[ARRAYIDX2]], align 4
31; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
32; CHECK:       for.body:
33; CHECK-NEXT:    [[TMP3:%.*]] = phi float [ [[TMP0]], [[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:    [[G_031:%.*]] = phi float [ [[TMP1]], [[ENTRY]] ], [ [[ADD9:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ]
37; CHECK-NEXT:    [[R_030:%.*]] = phi float [ [[TMP0]], [[ENTRY]] ], [ [[ADD4:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ]
38; CHECK-NEXT:    [[MUL:%.*]] = fmul float [[TMP3]], 7.000000e+00
39; CHECK-NEXT:    [[ADD4]] = fadd float [[R_030]], [[MUL]]
40; CHECK-NEXT:    [[TMP4:%.*]] = add nsw i64 [[INDVARS_IV]], 1
41; CHECK-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
42; CHECK-NEXT:    [[TMP5:%.*]] = load float, float* [[ARRAYIDX7]], align 4
43; CHECK-NEXT:    [[MUL8:%.*]] = fmul float [[TMP5]], 8.000000e+00
44; CHECK-NEXT:    [[ADD9]] = fadd float [[G_031]], [[MUL8]]
45; CHECK-NEXT:    [[TMP6:%.*]] = add nsw i64 [[INDVARS_IV]], 2
46; CHECK-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP6]]
47; CHECK-NEXT:    [[TMP7:%.*]] = load float, float* [[ARRAYIDX12]], align 4
48; CHECK-NEXT:    [[MUL13:%.*]] = fmul float [[TMP7]], 9.000000e+00
49; CHECK-NEXT:    [[ADD14]] = fadd float [[B_032]], [[MUL13]]
50; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 3
51; CHECK-NEXT:    [[TMP8:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
52; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], 121
53; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY_FOR_BODY_CRIT_EDGE]], label [[FOR_END:%.*]]
54; CHECK:       for.body.for.body_crit_edge:
55; CHECK-NEXT:    [[ARRAYIDX3_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV_NEXT]]
56; CHECK-NEXT:    [[DOTPRE]] = load float, float* [[ARRAYIDX3_PHI_TRANS_INSERT]], align 4
57; CHECK-NEXT:    br label [[FOR_BODY]]
58; CHECK:       for.end:
59; CHECK-NEXT:    [[ADD16:%.*]] = fadd float [[ADD4]], [[ADD9]]
60; CHECK-NEXT:    [[ADD17:%.*]] = fadd float [[ADD16]], [[ADD14]]
61; CHECK-NEXT:    ret float [[ADD17]]
62;
63entry:
64  %0 = load float, float* %A, align 4
65  %arrayidx1 = getelementptr inbounds float, float* %A, i64 1
66  %1 = load float, float* %arrayidx1, align 4
67  %arrayidx2 = getelementptr inbounds float, float* %A, i64 2
68  %2 = load float, float* %arrayidx2, align 4
69  br label %for.body
70
71for.body:                                         ; preds = %for.body.for.body_crit_edge, %entry
72  %3 = phi float [ %0, %entry ], [ %.pre, %for.body.for.body_crit_edge ]
73  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body.for.body_crit_edge ]
74  %B.032 = phi float [ %2, %entry ], [ %add14, %for.body.for.body_crit_edge ]
75  %G.031 = phi float [ %1, %entry ], [ %add9, %for.body.for.body_crit_edge ]
76  %R.030 = phi float [ %0, %entry ], [ %add4, %for.body.for.body_crit_edge ]
77  %mul = fmul float %3, 7.000000e+00
78  %add4 = fadd float %R.030, %mul
79  %4 = add nsw i64 %indvars.iv, 1
80  %arrayidx7 = getelementptr inbounds float, float* %A, i64 %4
81  %5 = load float, float* %arrayidx7, align 4
82  %mul8 = fmul float %5, 8.000000e+00
83  %add9 = fadd float %G.031, %mul8
84  %6 = add nsw i64 %indvars.iv, 2
85  %arrayidx12 = getelementptr inbounds float, float* %A, i64 %6
86  %7 = load float, float* %arrayidx12, align 4
87  %mul13 = fmul float %7, 9.000000e+00
88  %add14 = fadd float %B.032, %mul13
89  %indvars.iv.next = add i64 %indvars.iv, 3
90  %8 = trunc i64 %indvars.iv.next to i32
91  %cmp = icmp slt i32 %8, 121
92  br i1 %cmp, label %for.body.for.body_crit_edge, label %for.end
93
94for.body.for.body_crit_edge:                      ; preds = %for.body
95  %arrayidx3.phi.trans.insert = getelementptr inbounds float, float* %A, i64 %indvars.iv.next
96  %.pre = load float, float* %arrayidx3.phi.trans.insert, align 4
97  br label %for.body
98
99for.end:                                          ; preds = %for.body
100  %add16 = fadd float %add4, %add9
101  %add17 = fadd float %add16, %add14
102  ret float %add17
103}
104
105