xref: /llvm-project/llvm/test/CodeGen/X86/avx512-fma-commute.ll (revision 2f448bf509432c1a19ec46ab8cbc7353c03c6280) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -disable-peephole -mtriple=x86_64-apple-darwin -mattr=avx512f | FileCheck %s
3
4declare <4 x float> @llvm.x86.avx512.mask3.vfmadd.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32)
5declare <2 x double> @llvm.x86.avx512.mask3.vfmadd.sd(<2 x double>, <2 x double>, <2 x double>, i8, i32)
6declare <4 x float> @llvm.x86.avx512.mask3.vfmsub.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32)
7declare <2 x double> @llvm.x86.avx512.mask3.vfmsub.sd(<2 x double>, <2 x double>, <2 x double>, i8, i32)
8
9define <4 x float> @test_int_x86_avx512_mask3_vfmadd_ss_load0(ptr %x0ptr, <4 x float> %x1, <4 x float> %x2) {
10; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_ss_load0:
11; CHECK:       ## %bb.0:
12; CHECK-NEXT:    vfmadd231ss {{.*#+}} xmm1 = (xmm0 * mem) + xmm1
13; CHECK-NEXT:    vmovaps %xmm1, %xmm0
14; CHECK-NEXT:    retq
15  %x0 = load <4 x float>, ptr %x0ptr
16  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1, i32 4)
17  ret <4 x float> %res
18}
19
20define <4 x float> @test_int_x86_avx512_mask3_vfmadd_ss_load1(<4 x float> %x0, ptr %x1ptr, <4 x float> %x2){
21; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_ss_load1:
22; CHECK:       ## %bb.0:
23; CHECK-NEXT:    vfmadd231ss {{.*#+}} xmm1 = (xmm0 * mem) + xmm1
24; CHECK-NEXT:    vmovaps %xmm1, %xmm0
25; CHECK-NEXT:    retq
26  %x1 = load <4 x float>, ptr %x1ptr
27  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1, i32 4)
28  ret <4 x float> %res
29}
30
31define <2 x double> @test_int_x86_avx512_mask3_vfmadd_sd_load0(ptr %x0ptr, <2 x double> %x1, <2 x double> %x2) {
32; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_sd_load0:
33; CHECK:       ## %bb.0:
34; CHECK-NEXT:    vfmadd231sd {{.*#+}} xmm1 = (xmm0 * mem) + xmm1
35; CHECK-NEXT:    vmovapd %xmm1, %xmm0
36; CHECK-NEXT:    retq
37  %x0 = load <2 x double>, ptr %x0ptr
38  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmadd.sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1, i32 4)
39  ret <2 x double> %res
40}
41
42define <2 x double> @test_int_x86_avx512_mask3_vfmadd_sd_load1(<2 x double> %x0, ptr %x1ptr, <2 x double> %x2){
43; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_sd_load1:
44; CHECK:       ## %bb.0:
45; CHECK-NEXT:    vfmadd231sd {{.*#+}} xmm1 = (xmm0 * mem) + xmm1
46; CHECK-NEXT:    vmovapd %xmm1, %xmm0
47; CHECK-NEXT:    retq
48  %x1 = load <2 x double>, ptr %x1ptr
49  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmadd.sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1, i32 4)
50  ret <2 x double> %res
51}
52
53define <4 x float> @test_int_x86_avx512_mask3_vfmsub_ss_load0(ptr %x0ptr, <4 x float> %x1, <4 x float> %x2) {
54; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_ss_load0:
55; CHECK:       ## %bb.0:
56; CHECK-NEXT:    vfmsub231ss {{.*#+}} xmm1 = (xmm0 * mem) - xmm1
57; CHECK-NEXT:    vmovaps %xmm1, %xmm0
58; CHECK-NEXT:    retq
59  %x0 = load <4 x float>, ptr %x0ptr
60  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmsub.ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1, i32 4)
61  ret <4 x float> %res
62}
63
64define <4 x float> @test_int_x86_avx512_mask3_vfmsub_ss_load1(<4 x float> %x0, ptr %x1ptr, <4 x float> %x2){
65; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_ss_load1:
66; CHECK:       ## %bb.0:
67; CHECK-NEXT:    vfmsub231ss {{.*#+}} xmm1 = (xmm0 * mem) - xmm1
68; CHECK-NEXT:    vmovaps %xmm1, %xmm0
69; CHECK-NEXT:    retq
70  %x1 = load <4 x float>, ptr %x1ptr
71  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmsub.ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1, i32 4)
72  ret <4 x float> %res
73}
74
75define <2 x double> @test_int_x86_avx512_mask3_vfmsub_sd_load0(ptr %x0ptr, <2 x double> %x1, <2 x double> %x2) {
76; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_sd_load0:
77; CHECK:       ## %bb.0:
78; CHECK-NEXT:    vfmsub231sd {{.*#+}} xmm1 = (xmm0 * mem) - xmm1
79; CHECK-NEXT:    vmovapd %xmm1, %xmm0
80; CHECK-NEXT:    retq
81  %x0 = load <2 x double>, ptr %x0ptr
82  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmsub.sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1, i32 4)
83  ret <2 x double> %res
84}
85
86define <2 x double> @test_int_x86_avx512_mask3_vfmsub_sd_load1(<2 x double> %x0, ptr %x1ptr, <2 x double> %x2){
87; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_sd_load1:
88; CHECK:       ## %bb.0:
89; CHECK-NEXT:    vfmsub231sd {{.*#+}} xmm1 = (xmm0 * mem) - xmm1
90; CHECK-NEXT:    vmovapd %xmm1, %xmm0
91; CHECK-NEXT:    retq
92  %x1 = load <2 x double>, ptr %x1ptr
93  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmsub.sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1, i32 4)
94  ret <2 x double> %res
95}
96