1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=i686-darwin -mattr=+mmx,+sse2 | FileCheck %s --check-prefix=X32
3; RUN: llc < %s -mtriple=x86_64-darwin -mattr=+mmx,+sse4.1 | FileCheck %s --check-prefix=X64
4
5; This is not an MMX operation; promoted to xmm.
6define x86_mmx @t0(i32 %A) nounwind {
7; X32-LABEL: t0:
8; X32:       ## %bb.0:
9; X32-NEXT:    movd {{[0-9]+}}(%esp), %mm1
10; X32-NEXT:    pxor %mm0, %mm0
11; X32-NEXT:    punpckldq %mm1, %mm0 ## mm0 = mm0[0],mm1[0]
12; X32-NEXT:    retl
13;
14; X64-LABEL: t0:
15; X64:       ## %bb.0:
16; X64-NEXT:    movd %edi, %xmm0
17; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,1,1]
18; X64-NEXT:    retq
19  %tmp3 = insertelement <2 x i32> < i32 0, i32 undef >, i32 %A, i32 1
20  %tmp4 = bitcast <2 x i32> %tmp3 to x86_mmx
21  ret x86_mmx %tmp4
22}
23
24define <8 x i8> @t1(i8 zeroext %x) nounwind {
25; X32-LABEL: t1:
26; X32:       ## %bb.0:
27; X32-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
28; X32-NEXT:    retl
29;
30; X64-LABEL: t1:
31; X64:       ## %bb.0:
32; X64-NEXT:    movd %edi, %xmm0
33; X64-NEXT:    retq
34  %r = insertelement <8 x i8> undef, i8 %x, i32 0
35  ret <8 x i8> %r
36}
37
38; PR2574
39define <2 x float> @t2(<2 x float> %a0) {
40; X32-LABEL: t2:
41; X32:       ## %bb.0:
42; X32-NEXT:    xorps %xmm0, %xmm0
43; X32-NEXT:    retl
44;
45; X64-LABEL: t2:
46; X64:       ## %bb.0:
47; X64-NEXT:    xorps %xmm0, %xmm0
48; X64-NEXT:    retq
49  %v1 = insertelement <2 x float> %a0, float 0.000000e+00, i32 0
50  %v2 = insertelement <2 x float> %v1, float 0.000000e+00, i32 1
51  ret <2 x float> %v2
52}
53
54@g0 = external global i16
55@g1 = external global <4 x i16>
56
57; PR2562
58define void @t3() {
59; X32-LABEL: t3:
60; X32:       ## %bb.0:
61; X32-NEXT:    movl L_g0$non_lazy_ptr, %eax
62; X32-NEXT:    movl L_g1$non_lazy_ptr, %ecx
63; X32-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
64; X32-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
65; X32-NEXT:    movzwl (%eax), %eax
66; X32-NEXT:    movd %eax, %xmm1
67; X32-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
68; X32-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
69; X32-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
70; X32-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
71; X32-NEXT:    movq %xmm0, (%ecx)
72; X32-NEXT:    retl
73;
74; X64-LABEL: t3:
75; X64:       ## %bb.0:
76; X64-NEXT:    movq _g0@{{.*}}(%rip), %rax
77; X64-NEXT:    movq _g1@{{.*}}(%rip), %rcx
78; X64-NEXT:    pmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
79; X64-NEXT:    movzwl (%rax), %eax
80; X64-NEXT:    pinsrd $0, %eax, %xmm0
81; X64-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
82; X64-NEXT:    movq %xmm0, (%rcx)
83; X64-NEXT:    retq
84  load i16, i16* @g0
85  load <4 x i16>, <4 x i16>* @g1
86  insertelement <4 x i16> %2, i16 %1, i32 0
87  store <4 x i16> %3, <4 x i16>* @g1
88  ret void
89}
90