• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X32
3; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X64
4
5define void @t1(float* %R, <4 x float>* %P1) nounwind {
6; X32-LABEL: t1:
7; X32:       # BB#0:
8; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
9; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
10; X32-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
11; X32-NEXT:    movss %xmm0, (%eax)
12; X32-NEXT:    retl
13;
14; X64-LABEL: t1:
15; X64:       # BB#0:
16; X64-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
17; X64-NEXT:    movss %xmm0, (%rdi)
18; X64-NEXT:    retq
19  %X = load <4 x float>, <4 x float>* %P1
20  %tmp = extractelement <4 x float> %X, i32 3
21  store float %tmp, float* %R
22  ret void
23}
24
25define float @t2(<4 x float>* %P1) nounwind {
26; X32-LABEL: t2:
27; X32:       # BB#0:
28; X32-NEXT:    pushl %eax
29; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
30; X32-NEXT:    movddup {{.*#+}} xmm0 = mem[0,0]
31; X32-NEXT:    movss %xmm0, (%esp)
32; X32-NEXT:    flds (%esp)
33; X32-NEXT:    popl %eax
34; X32-NEXT:    retl
35;
36; X64-LABEL: t2:
37; X64:       # BB#0:
38; X64-NEXT:    movddup {{.*#+}} xmm0 = mem[0,0]
39; X64-NEXT:    retq
40  %X = load <4 x float>, <4 x float>* %P1
41  %tmp = extractelement <4 x float> %X, i32 2
42  ret float %tmp
43}
44
45define void @t3(i32* %R, <4 x i32>* %P1) nounwind {
46; X32-LABEL: t3:
47; X32:       # BB#0:
48; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
49; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
50; X32-NEXT:    movl 12(%ecx), %ecx
51; X32-NEXT:    movl %ecx, (%eax)
52; X32-NEXT:    retl
53;
54; X64-LABEL: t3:
55; X64:       # BB#0:
56; X64-NEXT:    movl 12(%rsi), %eax
57; X64-NEXT:    movl %eax, (%rdi)
58; X64-NEXT:    retq
59  %X = load <4 x i32>, <4 x i32>* %P1
60  %tmp = extractelement <4 x i32> %X, i32 3
61  store i32 %tmp, i32* %R
62  ret void
63}
64
65define i32 @t4(<4 x i32>* %P1) nounwind {
66; X32-LABEL: t4:
67; X32:       # BB#0:
68; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
69; X32-NEXT:    movl 12(%eax), %eax
70; X32-NEXT:    retl
71;
72; X64-LABEL: t4:
73; X64:       # BB#0:
74; X64-NEXT:    movl 12(%rdi), %eax
75; X64-NEXT:    retq
76  %X = load <4 x i32>, <4 x i32>* %P1
77  %tmp = extractelement <4 x i32> %X, i32 3
78  ret i32 %tmp
79}
80