1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
3 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2
5 ; Prefer a blend instruction to a vinsert128 instruction because blends
6 ; are simpler (no lane changes) and therefore will have equal or better
9 define <8 x float> @castA(<4 x float> %m) nounwind uwtable readnone ssp {
12 ; AVX-NEXT: vxorps %ymm1, %ymm1, %ymm1
13 ; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
15 %shuffle.i = shufflevector <4 x float> %m, <4 x float> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 4, i32 4, i32 4>
16 ret <8 x float> %shuffle.i
19 define <4 x double> @castB(<2 x double> %m) nounwind uwtable readnone ssp {
22 ; AVX-NEXT: vxorpd %ymm1, %ymm1, %ymm1
23 ; AVX-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
25 %shuffle.i = shufflevector <2 x double> %m, <2 x double> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 2, i32 2>
26 ret <4 x double> %shuffle.i
29 ; AVX2 is needed for integer types.
31 define <4 x i64> @castC(<2 x i64> %m) nounwind uwtable readnone ssp {
34 ; AVX1-NEXT: vxorps %ymm1, %ymm1, %ymm1
35 ; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
40 ; AVX2-NEXT: vpxor %ymm1, %ymm1, %ymm1
41 ; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
43 %shuffle.i = shufflevector <2 x i64> %m, <2 x i64> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 2, i32 2>
44 ret <4 x i64> %shuffle.i
47 ; The next three tests don't need any shuffling. There may or may not be a
48 ; vzeroupper before the return, so just check for the absence of shuffles.
50 define <4 x float> @castD(<8 x float> %m) nounwind uwtable readnone ssp {
53 ; AVX-NEXT: vzeroupper
55 %shuffle.i = shufflevector <8 x float> %m, <8 x float> %m, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
56 ret <4 x float> %shuffle.i
59 define <2 x i64> @castE(<4 x i64> %m) nounwind uwtable readnone ssp {
62 ; AVX-NEXT: vzeroupper
64 %shuffle.i = shufflevector <4 x i64> %m, <4 x i64> %m, <2 x i32> <i32 0, i32 1>
65 ret <2 x i64> %shuffle.i
68 define <2 x double> @castF(<4 x double> %m) nounwind uwtable readnone ssp {
71 ; AVX-NEXT: vzeroupper
73 %shuffle.i = shufflevector <4 x double> %m, <4 x double> %m, <2 x i32> <i32 0, i32 1>
74 ret <2 x double> %shuffle.i