1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -verify-machineinstrs -mtriple=powerpc64-linux-gnu -mcpu=pwr8 -mattr=+vsx < %s | FileCheck %s
3
4define <4 x float> @bar(float* %p, float* %q) {
5; CHECK-LABEL: bar:
6; CHECK:       # %bb.0:
7; CHECK-NEXT:    lxvw4x 0, 0, 3
8; CHECK-NEXT:    lxvw4x 1, 0, 4
9; CHECK-NEXT:    li 5, 16
10; CHECK-NEXT:    lxvw4x 2, 3, 5
11; CHECK-NEXT:    lxvw4x 3, 4, 5
12; CHECK-NEXT:    li 5, 32
13; CHECK-NEXT:    lxvw4x 4, 4, 5
14; CHECK-NEXT:    xvsubsp 0, 1, 0
15; CHECK-NEXT:    lxvw4x 1, 3, 5
16; CHECK-NEXT:    xvsubsp 34, 3, 2
17; CHECK-NEXT:    xvsubsp 35, 4, 1
18; CHECK-NEXT:    xxsldwi 36, 0, 0, 1
19; CHECK-NEXT:    vmrglw 2, 4, 2
20; CHECK-NEXT:    vmrghw 3, 3, 3
21; CHECK-NEXT:    xxsldwi 0, 36, 34, 3
22; CHECK-NEXT:    xxsldwi 0, 35, 0, 3
23; CHECK-NEXT:    xxsldwi 34, 0, 0, 1
24; CHECK-NEXT:    blr
25  %1 = bitcast float* %p to <12 x float>*
26  %2 = bitcast float* %q to <12 x float>*
27  %3 = load <12 x float>, <12 x float>* %1, align 16
28  %4 = load <12 x float>, <12 x float>* %2, align 16
29  %5 = fsub <12 x float> %4, %3
30  %6 = shufflevector <12 x float> %5, <12 x float> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
31  ret <4 x float>  %6
32}
33