1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -march=x86-64 -mcpu=nehalem | FileCheck %s
3 define <4 x float> @a(<4 x float>* %y) nounwind {
6 ; CHECK-NEXT: movdqu (%rdi), %xmm0
7 ; CHECK-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,2,1,0]
9 %x = load <4 x float>* %y, align 4
10 %a = extractelement <4 x float> %x, i32 0
11 %b = extractelement <4 x float> %x, i32 1
12 %c = extractelement <4 x float> %x, i32 2
13 %d = extractelement <4 x float> %x, i32 3
14 %p = insertelement <4 x float> undef, float %d, i32 0
15 %q = insertelement <4 x float> %p, float %c, i32 1
16 %r = insertelement <4 x float> %q, float %b, i32 2
17 %s = insertelement <4 x float> %r, float %a, i32 3
21 define <4 x float> @b(<4 x float>* %y, <4 x float> %z) nounwind {
24 ; CHECK-NEXT: movups (%rdi), %xmm1
25 ; CHECK-NEXT: unpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
27 %x = load <4 x float>* %y, align 4
28 %a = extractelement <4 x float> %x, i32 2
29 %b = extractelement <4 x float> %x, i32 3
30 %c = extractelement <4 x float> %z, i32 2
31 %d = extractelement <4 x float> %z, i32 3
32 %p = insertelement <4 x float> undef, float %c, i32 0
33 %q = insertelement <4 x float> %p, float %a, i32 1
34 %r = insertelement <4 x float> %q, float %d, i32 2
35 %s = insertelement <4 x float> %r, float %b, i32 3
39 define <2 x double> @c(<2 x double>* %y) nounwind {
42 ; CHECK-NEXT: movupd (%rdi), %xmm0
43 ; CHECK-NEXT: shufpd {{.*#+}} xmm0 = xmm0[1,0]
45 %x = load <2 x double>* %y, align 8
46 %a = extractelement <2 x double> %x, i32 0
47 %c = extractelement <2 x double> %x, i32 1
48 %p = insertelement <2 x double> undef, double %c, i32 0
49 %r = insertelement <2 x double> %p, double %a, i32 1
53 define <2 x double> @d(<2 x double>* %y, <2 x double> %z) nounwind {
56 ; CHECK-NEXT: movupd (%rdi), %xmm1
57 ; CHECK-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
59 %x = load <2 x double>* %y, align 8
60 %a = extractelement <2 x double> %x, i32 1
61 %c = extractelement <2 x double> %z, i32 1
62 %p = insertelement <2 x double> undef, double %c, i32 0
63 %r = insertelement <2 x double> %p, double %a, i32 1