xref: /llvm-project/llvm/test/CodeGen/X86/coalesce_commute_movsd.ll (revision 1deca50ab1479d4721a9c01a6c5b82a39258abe5)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -verify-machineinstrs -verify-coalescing -mattr=+sse2 | FileCheck %s --check-prefix=SSE2
3; RUN: llc < %s -mtriple=x86_64-unknown-unknown -verify-machineinstrs -verify-coalescing -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE41
4; RUN: llc < %s -mtriple=x86_64-unknown-unknown -verify-machineinstrs -verify-coalescing -mattr=+avx | FileCheck %s --check-prefix=AVX
5; RUN: llc < %s -mtriple=x86_64-unknown-unknown -verify-machineinstrs -verify-coalescing -mattr=+avx512f | FileCheck %s --check-prefix=AVX512
6
7; PR30607
8
9define <2 x double> @insert_f64(double %a0, <2 x double> %a1) {
10; SSE2-LABEL: insert_f64:
11; SSE2:       # %bb.0:
12; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
13; SSE2-NEXT:    retq
14;
15; SSE41-LABEL: insert_f64:
16; SSE41:       # %bb.0:
17; SSE41-NEXT:    blendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
18; SSE41-NEXT:    retq
19;
20; AVX-LABEL: insert_f64:
21; AVX:       # %bb.0:
22; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
23; AVX-NEXT:    retq
24;
25; AVX512-LABEL: insert_f64:
26; AVX512:       # %bb.0:
27; AVX512-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
28; AVX512-NEXT:    retq
29 %1 = insertelement <2 x double> %a1, double %a0, i32 0
30 ret <2 x double> %1
31}
32
33define <4 x float> @insert_f32(float %a0, <4 x float> %a1) {
34; SSE2-LABEL: insert_f32:
35; SSE2:       # %bb.0:
36; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
37; SSE2-NEXT:    movaps %xmm1, %xmm0
38; SSE2-NEXT:    retq
39;
40; SSE41-LABEL: insert_f32:
41; SSE41:       # %bb.0:
42; SSE41-NEXT:    blendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
43; SSE41-NEXT:    retq
44;
45; AVX-LABEL: insert_f32:
46; AVX:       # %bb.0:
47; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
48; AVX-NEXT:    retq
49;
50; AVX512-LABEL: insert_f32:
51; AVX512:       # %bb.0:
52; AVX512-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
53; AVX512-NEXT:    retq
54 %1 = insertelement <4 x float> %a1, float %a0, i32 0
55 ret <4 x float> %1
56}
57