1; RUN: opt -mtriple=x86_64-unknown-linux-gnu -load-store-vectorizer -mcpu haswell -S -o - %s | FileCheck --check-prefix=CHECK-HSW %s
2; RUN: opt -mtriple=x86_64-unknown-linux-gnu -load-store-vectorizer -mcpu knl -S -o - %s | FileCheck --check-prefix=CHECK-KNL %s
3
4define <8 x double> @loadwidth_insert_extract(double* %ptr) {
5    %a = bitcast double* %ptr to <2 x double> *
6    %b = getelementptr <2 x double>, <2 x double>* %a, i32 1
7    %c = getelementptr <2 x double>, <2 x double>* %a, i32 2
8    %d = getelementptr <2 x double>, <2 x double>* %a, i32 3
9; CHECK-HSW: load <4 x double>
10; CHECK-HSW: load <4 x double>
11; CHECK-HSW-NOT: load
12; CHECK-KNL: load <8 x double>
13; CHECK-KNL-NOT: load
14    %la = load <2 x double>, <2 x double> *%a
15    %lb = load <2 x double>, <2 x double> *%b
16    %lc = load <2 x double>, <2 x double> *%c
17    %ld = load <2 x double>, <2 x double> *%d
18    ; Scalarize everything - Explicitly not a shufflevector to test this code
19    ; path in the LSV
20    %v1 = extractelement <2 x double> %la, i32 0
21    %v2 = extractelement <2 x double> %la, i32 1
22    %v3 = extractelement <2 x double> %lb, i32 0
23    %v4 = extractelement <2 x double> %lb, i32 1
24    %v5 = extractelement <2 x double> %lc, i32 0
25    %v6 = extractelement <2 x double> %lc, i32 1
26    %v7 = extractelement <2 x double> %ld, i32 0
27    %v8 = extractelement <2 x double> %ld, i32 1
28    ; Make a vector again
29    %i1 = insertelement <8 x double> undef, double %v1, i32 0
30    %i2 = insertelement <8 x double> %i1, double %v2, i32 1
31    %i3 = insertelement <8 x double> %i2, double %v3, i32 2
32    %i4 = insertelement <8 x double> %i3, double %v4, i32 3
33    %i5 = insertelement <8 x double> %i4, double %v5, i32 4
34    %i6 = insertelement <8 x double> %i5, double %v6, i32 5
35    %i7 = insertelement <8 x double> %i6, double %v7, i32 6
36    %i8 = insertelement <8 x double> %i7, double %v8, i32 7
37    ret <8 x double> %i8
38}
39