[x86] Teach the decomposed shuffle/blend lowering to use an early blend

[oota-llvm.git] / test / CodeGen / X86 / vector-shuffle-512-v8.ll

diff --git a/test/CodeGen/X86/vector-shuffle-512-v8.ll b/test/CodeGen/X86/vector-shuffle-512-v8.ll
index 09dde3132dded06ab8b2fd8b3887d126036b8fb1..f9e131b40397b55dacdc25f3a1aef3d9fa192892 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -1,13 +1,12 @@
-; RUN: llc < %s -mcpu=x86-64 -mattr=+avx512f -x86-experimental-vector-shuffle-lowering | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: llc < %s -mcpu=x86-64 -mattr=+avx512bw -x86-experimental-vector-shuffle-lowering | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 --check-prefix=AVX512BW
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx512f -x86-experimental-vector-shuffle-legality | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 --check-prefix=AVX512F
+; RUN: llc < %s -mcpu=x86-64 -mattr=+avx512bw -x86-experimental-vector-shuffle-legality | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 --check-prefix=AVX512BW
 
 target triple = "x86_64-unknown-unknown"
 
 define <8 x double> @shuffle_v8f64_00000000(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00000000:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,0]
-; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
+; ALL-NEXT:    vbroadcastsd %xmm0, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <8 x double> %shuffle
@@ -16,7 +15,7 @@ define <8 x double> @shuffle_v8f64_00000000(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_00000010(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00000010:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm1
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -27,7 +26,7 @@ define <8 x double> @shuffle_v8f64_00000010(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_00000200(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00000200:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm1
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -38,7 +37,7 @@ define <8 x double> @shuffle_v8f64_00000200(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_00003000(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00003000:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm1
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,0,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -50,10 +49,9 @@ define <8 x double> @shuffle_v8f64_00040000(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00040000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,0]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm0[0,0,0,3]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm1, %ymm1
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
+; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
@@ -64,10 +62,9 @@ define <8 x double> @shuffle_v8f64_00500000(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00500000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,1,1,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm0[0,0,2,0]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2,3]
+; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,1,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -78,10 +75,9 @@ define <8 x double> @shuffle_v8f64_06000000(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_06000000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm0[0,1,0,0]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2],ymm0[3]
+; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -92,9 +88,9 @@ define <8 x double> @shuffle_v8f64_70000000(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_70000000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,1,2,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,0]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm0[1,2,3]
+; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,0,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -104,10 +100,10 @@ define <8 x double> @shuffle_v8f64_70000000(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_01014545(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_01014545:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,1,0,1]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
+; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm1, %ymm1
+; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; ALL-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
   ret <8 x double> %shuffle
@@ -127,7 +123,7 @@ define <8 x double> @shuffle_v8f64_00112233(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_00001111(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00001111:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm1
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,1,1]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -151,8 +147,8 @@ define <8 x double> @shuffle_v8f64_81a3c5e7(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_08080808(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_08080808:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,0]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; ALL-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -163,13 +159,13 @@ define <8 x double> @shuffle_v8f64_08080808(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_08084c4c(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_08084c4c:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
-; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm3
-; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,0]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
+; ALL-NEXT:    vinsertf128 $1, %xmm2, %ymm2, %ymm2
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
+; ALL-NEXT:    vbroadcastsd %xmm3, %ymm3
+; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[3]
+; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; ALL-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -182,9 +178,9 @@ define <8 x double> @shuffle_v8f64_8823cc67(<8 x double> %a, <8 x double> %b) {
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
 ; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[0,0,2,2]
+; ALL-NEXT:    vbroadcastsd %xmm3, %ymm3
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,2,2]
+; ALL-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -195,15 +191,13 @@ define <8 x double> @shuffle_v8f64_8823cc67(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_9832dc76(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_9832dc76:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
+; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm1[0,1],ymm0[2,3]
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[1,0,3,2]
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm1
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 9, i32 8, i32 3, i32 2, i32 13, i32 12, i32 7, i32 6>
   ret <8 x double> %shuffle
@@ -212,15 +206,13 @@ define <8 x double> @shuffle_v8f64_9832dc76(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_9810dc54(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_9810dc54:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,1,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,0]
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm2
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm2[1,0,3,2]
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm1
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
+; ALL-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
   ret <8 x double> %shuffle
@@ -278,12 +270,11 @@ define <8 x double> @shuffle_v8f64_08192a3b(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_08991abb(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_08991abb:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm0[1,0,2,2]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm1[0,2,3,3]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,1,1]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
-; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm1[0,0,1,1]
+; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm0[0],ymm2[1,2,3]
+; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3]
+; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,2,3,3]
+; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
   ret <8 x double> %shuffle
@@ -415,9 +406,9 @@ define <8 x double> @shuffle_v8f64_00234467(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_00224466(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00224466:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,2]
+; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,2]
+; ALL-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
@@ -513,7 +504,7 @@ define <8 x double> @shuffle_v8f64_10004444(<8 x double> %a, <8 x double> %b) {
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[1,0,0,0]
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
@@ -570,7 +561,7 @@ define <8 x double> @shuffle_v8f64_00236744(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_00226644(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00226644:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,2]
+; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
@@ -626,7 +617,7 @@ define <8 x double> @shuffle_v8f64_01235466(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_002u6u44(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_002u6u44:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,2]
+; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
@@ -638,7 +629,7 @@ define <8 x double> @shuffle_v8f64_002u6u44(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_00uu66uu(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00uu66uu:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,2]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm1
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
@@ -684,7 +675,7 @@ define <8 x double> @shuffle_v8f64_uuu3uu66(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_uuu3uu66:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,2,2]
+; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm1[0,0,2,2]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 3, i32 undef, i32 undef, i32 6, i32 6>
@@ -695,15 +686,13 @@ define <8 x double> @shuffle_v8f64_c348cda0(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_c348cda0:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm0[0,3,2,3]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2],ymm3[3]
+; ALL-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[0,1],ymm2[0,1]
 ; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
-; ALL-NEXT:    vpermpd {{.*#+}} ymm4 = ymm1[0,1,2,0]
+; ALL-NEXT:    vbroadcastsd %xmm1, %ymm4
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm4 = ymm3[0,1,2],ymm4[3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0],ymm2[1,2],ymm4[3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2],ymm3[3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,0]
+; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
@@ -715,10 +704,9 @@ define <8 x double> @shuffle_v8f64_f511235a(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_f511235a:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
-; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm2[0,1,1,3]
-; ALL-NEXT:    vpermpd {{.*#+}} ymm4 = ymm0[2,3,2,3]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3]
-; ALL-NEXT:    vpermilpd {{.*#+}} ymm4 = ymm1[0,0,2,2]
+; ALL-NEXT:    vblendpd {{.*#+}} ymm3 = ymm0[0],ymm2[1],ymm0[2,3]
+; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,3,1,3]
+; ALL-NEXT:    vmovddup {{.*#+}} ymm4 = ymm1[0,0,2,2]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3]
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,1]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
@@ -734,8 +722,7 @@ define <8 x double> @shuffle_v8f64_f511235a(<8 x double> %a, <8 x double> %b) {
 define <8 x i64> @shuffle_v8i64_00000000(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00000000:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,0]
-; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; ALL-NEXT:    vpbroadcastq %xmm0, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <8 x i64> %shuffle
@@ -744,7 +731,7 @@ define <8 x i64> @shuffle_v8i64_00000000(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_00000010(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00000010:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm1
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,1,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -755,7 +742,7 @@ define <8 x i64> @shuffle_v8i64_00000010(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_00000200(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00000200:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm1
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -766,7 +753,7 @@ define <8 x i64> @shuffle_v8i64_00000200(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_00003000(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00003000:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm1
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,0,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -778,10 +765,9 @@ define <8 x i64> @shuffle_v8i64_00040000(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00040000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
-; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[0,0,0,3]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm0
+; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
@@ -792,10 +778,9 @@ define <8 x i64> @shuffle_v8i64_00500000(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00500000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,1,1,3]
-; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[0,0,2,0]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
+; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -806,10 +791,9 @@ define <8 x i64> @shuffle_v8i64_06000000(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_06000000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[0,1,0,0]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -820,9 +804,9 @@ define <8 x i64> @shuffle_v8i64_70000000(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_70000000:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[3,1,2,3]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,0]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[3,0,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -832,10 +816,10 @@ define <8 x i64> @shuffle_v8i64_70000000(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_01014545(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_01014545:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,1,0,1]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
-; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; ALL-NEXT:    vinserti128 $1, %xmm1, %ymm1, %ymm1
+; ALL-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; ALL-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 4, i32 5, i32 4, i32 5>
   ret <8 x i64> %shuffle
@@ -855,7 +839,7 @@ define <8 x i64> @shuffle_v8i64_00112233(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_00001111(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00001111:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm1
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,1,1,1]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -879,8 +863,8 @@ define <8 x i64> @shuffle_v8i64_81a3c5e7(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_08080808(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_08080808:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,0]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; ALL-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -891,13 +875,13 @@ define <8 x i64> @shuffle_v8i64_08080808(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_08084c4c(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_08084c4c:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,0,2,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm3
-; ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,0]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
+; ALL-NEXT:    vinserti128 $1, %xmm2, %ymm2, %ymm2
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
+; ALL-NEXT:    vpbroadcastq %xmm3, %ymm3
+; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; ALL-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -910,9 +894,9 @@ define <8 x i64> @shuffle_v8i64_8823cc67(<8 x i64> %a, <8 x i64> %b) {
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
 ; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
-; ALL-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,1,0,1,4,5,4,5]
+; ALL-NEXT:    vpbroadcastq %xmm3, %ymm3
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,4,5]
+; ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -923,15 +907,13 @@ define <8 x i64> @shuffle_v8i64_8823cc67(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_9832dc76(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_9832dc76:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
+; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
-; ALL-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; ALL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 9, i32 8, i32 3, i32 2, i32 13, i32 12, i32 7, i32 6>
   ret <8 x i64> %shuffle
@@ -940,15 +922,13 @@ define <8 x i64> @shuffle_v8i64_9832dc76(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_9810dc54(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_9810dc54:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
-; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,1,1,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
-; ALL-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,1,0]
-; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm2
+; ALL-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm1
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; ALL-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
+; ALL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 9, i32 8, i32 1, i32 0, i32 13, i32 12, i32 5, i32 4>
   ret <8 x i64> %shuffle
@@ -1006,12 +986,11 @@ define <8 x i64> @shuffle_v8i64_08192a3b(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_08991abb(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_08991abb:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpshufd {{.*#+}} ymm2 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm1[0,2,3,3]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,1,1]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[0,0,1,1]
+; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,2,3,3]
+; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm2, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
   ret <8 x i64> %shuffle
@@ -1241,7 +1220,7 @@ define <8 x i64> @shuffle_v8i64_10004444(<8 x i64> %a, <8 x i64> %b) {
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[1,0,0,0]
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,0]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4>
@@ -1366,7 +1345,7 @@ define <8 x i64> @shuffle_v8i64_002u6u44(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_00uu66uu(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00uu66uu:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[0,1,0,1,4,5,4,5]
+; ALL-NEXT:    vpbroadcastq %xmm0, %ymm1
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
@@ -1422,18 +1401,53 @@ define <8 x i64> @shuffle_v8i64_uuu3uu66(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_6caa87e5(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_6caa87e5:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
-; ALL-NEXT:    vpermq {{.*#+}} ymm4 = ymm0[0,3,2,1]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
-; ALL-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,1,0,1,4,5,4,5]
+; ALL-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[0,1,0,1]
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
+; ALL-NEXT:    vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5],ymm2[6,7]
+; ALL-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,4,5]
-; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
-; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,1,2,3]
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; ALL-NEXT:    vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 6, i32 12, i32 10, i32 10, i32 8, i32 7, i32 14, i32 5>
   ret <8 x i64> %shuffle
 }
+
+define <8 x double> @shuffle_v8f64_082a4c6e(<8 x double> %a, <8 x double> %b) {
+; ALL-LABEL: shuffle_v8f64_082a4c6e:
+; ALL:       # BB#0:
+; ALL-NEXT:    vunpcklpd {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32><i32 0, i32 8, i32 2, i32 10, i32 4, i32 12, i32 6, i32 14>
+  ret <8 x double> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_082a4c6e(<8 x i64> %a, <8 x i64> %b) {
+; ALL-LABEL: shuffle_v8i64_082a4c6e:
+; ALL:       # BB#0:
+; ALL-NEXT:    vpunpcklqdq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32><i32 0, i32 8, i32 2, i32 10, i32 4, i32 12, i32 6, i32 14>
+  ret <8 x i64> %shuffle
+}
+
+define <8 x double> @shuffle_v8f64_193b5d7f(<8 x double> %a, <8 x double> %b) {
+; ALL-LABEL: shuffle_v8f64_193b5d7f:
+; ALL:       # BB#0:
+; ALL-NEXT:    vunpckhpd {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32><i32 1, i32 9, i32 3, i32 11, i32 5, i32 13, i32 7, i32 15>
+  ret <8 x double> %shuffle
+}
+
+define <8 x i64> @shuffle_v8i64_193b5d7f(<8 x i64> %a, <8 x i64> %b) {
+; ALL-LABEL: shuffle_v8i64_193b5d7f:
+; ALL:       # BB#0:
+; ALL-NEXT:    vpunpckhqdq {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
+; ALL-NEXT:    retq
+  %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32><i32 1, i32 9, i32 3, i32 11, i32 5, i32 13, i32 7, i32 15>
+  ret <8 x i64> %shuffle
+}