[X86][SSE] Add support for combining target shuffles to SHUFPD.

llvm-svn: 289407

3 files changed, 7 insertions, 10 deletions

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
index b5f8b661e29..9391b0c1f52 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -687,9 +687,8 @@ define <8 x float> @shuffle_v8f32_f511235a(<8 x float> %a, <8 x float> %b) {
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[3,1,2,2,7,5,6,6]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,1,1,4,5,5,5]
-; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
+; AVX1-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[3],ymm0[3]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
 ; AVX1-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
index d6541f68608..687098f9abf 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
@@ -858,14 +858,12 @@ define <8 x double> @combine_vpermi2var_8f64_identity(<8 x double> %x0, <8 x dou
 define <8 x double> @combine_vpermi2var_8f64_as_shufpd(<8 x double> %x0, <8 x double> %x1) {
 ; X32-LABEL: combine_vpermi2var_8f64_as_shufpd:
 ; X32:       # BB#0:
-; X32-NEXT:    vmovapd {{.*#+}} zmm2 = [1,0,8,0,2,0,10,0,5,0,13,0,6,0,15,0]
-; X32-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; X32-NEXT:    vshufpd {{.*#+}} zmm0 = zmm0[1],zmm1[0],zmm0[2],zmm1[2],zmm0[5],zmm1[5],zmm0[6],zmm1[7]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermi2var_8f64_as_shufpd:
 ; X64:       # BB#0:
-; X64-NEXT:    vmovapd {{.*#+}} zmm2 = [1,8,2,10,5,13,6,15]
-; X64-NEXT:    vpermt2pd %zmm1, %zmm2, %zmm0
+; X64-NEXT:    vshufpd {{.*#+}} zmm0 = zmm0[1],zmm1[0],zmm0[2],zmm1[2],zmm0[5],zmm1[5],zmm0[6],zmm1[7]
 ; X64-NEXT:    retq
   %1 = call <8 x double> @llvm.x86.avx512.mask.vpermi2var.pd.512(<8 x double> %x0, <8 x i64> <i64 1, i64 8, i64 2, i64 10, i64 5, i64 13, i64 6, i64 15>, <8 x double> %x1, i8 -1)
   ret <8 x double> %1
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-xop.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-xop.ll
index 17a2aa6a16d..4ba4d17c85b 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-xop.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-xop.ll
@@ -150,12 +150,12 @@ define <4 x float> @combine_vpermil2ps_blend_with_zero(<4 x float> %a0, <4 x flo
 define <2 x double> @combine_vpermil2pd_as_shufpd(<2 x double> %a0, <2 x double> %a1) {
 ; X32-LABEL: combine_vpermil2pd_as_shufpd:
 ; X32:       # BB#0:
-; X32-NEXT:    vpermil2pd {{.*#+}} xmm0 = xmm0[1],xmm1[0]
+; X32-NEXT:    vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermil2pd_as_shufpd:
 ; X64:       # BB#0:
-; X64-NEXT:    vpermil2pd {{.*#+}} xmm0 = xmm0[1],xmm1[0]
+; X64-NEXT:    vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0]
 ; X64-NEXT:    retq
   %res0 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a0, <2 x double> %a1, <2 x i64> <i64 2, i64 4>, i8 0)
   ret <2 x double> %res0
@@ -164,12 +164,12 @@ define <2 x double> @combine_vpermil2pd_as_shufpd(<2 x double> %a0, <2 x double>
 define <4 x double> @combine_vpermil2pd256_as_shufpd(<4 x double> %a0, <4 x double> %a1) {
 ; X32-LABEL: combine_vpermil2pd256_as_shufpd:
 ; X32:       # BB#0:
-; X32-NEXT:    vpermil2pd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
+; X32-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: combine_vpermil2pd256_as_shufpd:
 ; X64:       # BB#0:
-; X64-NEXT:    vpermil2pd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
+; X64-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
 ; X64-NEXT:    retq
   %res0 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %a1, <4 x i64> <i64 0, i64 4, i64 2, i64 7>, i8 0)
   ret <4 x double> %res0