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:    li 5, 16
8; CHECK-NEXT:    lxvw4x 2, 0, 3
9; CHECK-NEXT:    lxvw4x 3, 0, 4
10; CHECK-NEXT:    addis 6, 2, .LCPI0_0@toc@ha
11; CHECK-NEXT:    lxvw4x 0, 3, 5
12; CHECK-NEXT:    lxvw4x 1, 4, 5
13; CHECK-NEXT:    li 5, 32
14; CHECK-NEXT:    xvsubsp 35, 3, 2
15; CHECK-NEXT:    xvsubsp 34, 1, 0
16; CHECK-NEXT:    lxvw4x 0, 3, 5
17; CHECK-NEXT:    addi 3, 6, .LCPI0_0@toc@l
18; CHECK-NEXT:    lxvw4x 1, 4, 5
19; CHECK-NEXT:    lxvw4x 36, 0, 3
20; CHECK-NEXT:    addis 3, 2, .LCPI0_1@toc@ha
21; CHECK-NEXT:    addi 3, 3, .LCPI0_1@toc@l
22; CHECK-NEXT:    xvsubsp 37, 1, 0
23; CHECK-NEXT:    vperm 2, 3, 2, 4
24; CHECK-NEXT:    lxvw4x 35, 0, 3
25; CHECK-NEXT:    vperm 2, 2, 5, 3
26; CHECK-NEXT:    blr
27  %1 = bitcast float* %p to <12 x float>*
28  %2 = bitcast float* %q to <12 x float>*
29  %3 = load <12 x float>, <12 x float>* %1, align 16
30  %4 = load <12 x float>, <12 x float>* %2, align 16
31  %5 = fsub <12 x float> %4, %3
32  %6 = shufflevector <12 x float> %5, <12 x float> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
33  ret <4 x float>  %6
34}
35