xref: /llvm-project/llvm/test/CodeGen/X86/vec_extract-sse4.ll (revision eb523a4d272e81c8f7bf48da3923ed502f41c187)
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=X86
3; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X64
4
5define void @t1(ptr %R, ptr %P1) nounwind {
6; X86-LABEL: t1:
7; X86:       # %bb.0:
8; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
9; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
10; X86-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
11; X86-NEXT:    movss %xmm0, (%eax)
12; X86-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>, ptr %P1
20  %tmp = extractelement <4 x float> %X, i32 3
21  store float %tmp, ptr %R
22  ret void
23}
24
25define float @t2(ptr %P1) nounwind {
26; X86-LABEL: t2:
27; X86:       # %bb.0:
28; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
29; X86-NEXT:    flds 8(%eax)
30; X86-NEXT:    retl
31;
32; X64-LABEL: t2:
33; X64:       # %bb.0:
34; X64-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
35; X64-NEXT:    retq
36  %X = load <4 x float>, ptr %P1
37  %tmp = extractelement <4 x float> %X, i32 2
38  ret float %tmp
39}
40
41define void @t3(ptr %R, ptr %P1) nounwind {
42; X86-LABEL: t3:
43; X86:       # %bb.0:
44; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
45; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
46; X86-NEXT:    movl 12(%ecx), %ecx
47; X86-NEXT:    movl %ecx, (%eax)
48; X86-NEXT:    retl
49;
50; X64-LABEL: t3:
51; X64:       # %bb.0:
52; X64-NEXT:    movl 12(%rsi), %eax
53; X64-NEXT:    movl %eax, (%rdi)
54; X64-NEXT:    retq
55  %X = load <4 x i32>, ptr %P1
56  %tmp = extractelement <4 x i32> %X, i32 3
57  store i32 %tmp, ptr %R
58  ret void
59}
60
61define i32 @t4(ptr %P1) nounwind {
62; X86-LABEL: t4:
63; X86:       # %bb.0:
64; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
65; X86-NEXT:    movl 12(%eax), %eax
66; X86-NEXT:    retl
67;
68; X64-LABEL: t4:
69; X64:       # %bb.0:
70; X64-NEXT:    movl 12(%rdi), %eax
71; X64-NEXT:    retq
72  %X = load <4 x i32>, ptr %P1
73  %tmp = extractelement <4 x i32> %X, i32 3
74  ret i32 %tmp
75}
76