1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -basic-aa -slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
3
4target 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"
5target triple = "x86_64-apple-macosx10.8.0"
6
7; Simple 3-pair chain with loads and stores
8define void @test1(double* %a, double* %b, double* %c) {
9; CHECK-LABEL: @test1(
10; CHECK-NEXT:  entry:
11; CHECK-NEXT:    [[AGG_TMP_I_I_SROA_0:%.*]] = alloca [3 x double], align 16
12; CHECK-NEXT:    [[STORE1:%.*]] = getelementptr inbounds [3 x double], [3 x double]* [[AGG_TMP_I_I_SROA_0]], i64 0, i64 1
13; CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
14; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
15; CHECK-NEXT:    [[TMP2:%.*]] = bitcast double* [[B:%.*]] to <2 x double>*
16; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8
17; CHECK-NEXT:    [[TMP4:%.*]] = fmul <2 x double> [[TMP1]], [[TMP3]]
18; CHECK-NEXT:    [[TMP5:%.*]] = bitcast double* [[STORE1]] to <2 x double>*
19; CHECK-NEXT:    store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
20; CHECK-NEXT:    ret void
21;
22entry:
23  %agg.tmp.i.i.sroa.0 = alloca [3 x double], align 16
24  %i0 = load double, double* %a
25  %i1 = load double, double* %b
26  %mul = fmul double %i0, %i1
27  %store1 = getelementptr inbounds [3 x double], [3 x double]* %agg.tmp.i.i.sroa.0, i64 0, i64 1
28  %store2 = getelementptr inbounds [3 x double], [3 x double]* %agg.tmp.i.i.sroa.0, i64 0, i64 2
29  %arrayidx3 = getelementptr inbounds double, double* %a, i64 1
30  %i3 = load double, double* %arrayidx3, align 8
31  %arrayidx4 = getelementptr inbounds double, double* %b, i64 1
32  %i4 = load double, double* %arrayidx4, align 8
33  %mul5 = fmul double %i3, %i4
34  store double %mul, double* %store1
35  store double %mul5, double* %store2, align 16
36  ret void
37}
38
39; Float has 4 byte abi alignment on x86_64. We must use the alignment of the
40; value being loaded/stored not the alignment of the pointer type.
41
42define void @test2(float * %a, float * %b) {
43; CHECK-LABEL: @test2(
44; CHECK-NEXT:  entry:
45; CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A:%.*]] to <4 x float>*
46; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[TMP0]], align 4
47; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[B:%.*]] to <4 x float>*
48; CHECK-NEXT:    store <4 x float> [[TMP1]], <4 x float>* [[TMP2]], align 4
49; CHECK-NEXT:    ret void
50;
51entry:
52  %l0 = load float, float* %a
53  %a1 = getelementptr inbounds float, float* %a, i64 1
54  %l1 = load float, float* %a1
55  %a2 = getelementptr inbounds float, float* %a, i64 2
56  %l2 = load float, float* %a2
57  %a3 = getelementptr inbounds float, float* %a, i64 3
58  %l3 = load float, float* %a3
59  store float %l0, float* %b
60  %b1 = getelementptr inbounds float, float* %b, i64 1
61  store float %l1, float* %b1
62  %b2 = getelementptr inbounds float, float* %b, i64 2
63  store float %l2, float* %b2
64  %b3 = getelementptr inbounds float, float* %b, i64 3
65  store float %l3, float* %b3
66  ret void
67}
68