[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 f3c64da044f0e9dcf76ae4eedd52dff381fe257b..f9e131b40397b55dacdc25f3a1aef3d9fa192892 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-512-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-512-v8.ll
@@ -1,12 +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:    vbroadcastsd %xmm0, %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
@@ -48,7 +48,7 @@ define <8 x double> @shuffle_v8f64_00003000(<8 x double> %a, <8 x double> %b) {
 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:    vextractf64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
@@ -61,10 +61,10 @@ define <8 x double> @shuffle_v8f64_00040000(<8 x double> %a, <8 x double> %b) {
 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:    vextractf64x4 $1, %zmm0, %ymm1
+; 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:    vblendpd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2],ymm0[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 5, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -74,10 +74,10 @@ define <8 x double> @shuffle_v8f64_00500000(<8 x double> %a, <8 x double> %b) {
 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:    vextractf64x4 $1, %zmm0, %ymm1
+; 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:    vblendpd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2,3]
 ; 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 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -87,10 +87,10 @@ define <8 x double> @shuffle_v8f64_06000000(<8 x double> %a, <8 x double> %b) {
 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:    vextractf64x4 $1, %zmm0, %ymm1
+; 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:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm0[1,2,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -100,7 +100,7 @@ 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:    vextractf64x4 $1, %zmm0, %ymm1 
+; 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
@@ -134,8 +134,8 @@ define <8 x double> @shuffle_v8f64_00001111(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_81a3c5e7(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_81a3c5e7:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2 
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
@@ -159,9 +159,9 @@ 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, %zmm0, %ymm2 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
 ; ALL-NEXT:    vinsertf128 $1, %xmm2, %ymm2, %ymm2
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3 
+; 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
@@ -176,8 +176,8 @@ define <8 x double> @shuffle_v8f64_08084c4c(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_8823cc67(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_8823cc67:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2 
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
 ; ALL-NEXT:    vbroadcastsd %xmm3, %ymm3
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
 ; ALL-NEXT:    vbroadcastsd %xmm1, %ymm1
@@ -221,9 +221,9 @@ define <8 x double> @shuffle_v8f64_9810dc54(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_08194c5d(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_08194c5d:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm3 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm3
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,0,2,1]
@@ -238,9 +238,9 @@ define <8 x double> @shuffle_v8f64_08194c5d(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_2a3b6e7f(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_2a3b6e7f:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm3 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm3
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[2,1,3,3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
@@ -270,12 +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
@@ -284,7 +283,7 @@ define <8 x double> @shuffle_v8f64_08991abb(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_091b2d3f(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_091b2d3f:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm3 = ymm0[2,1,3,3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2],ymm2[3]
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,3]
@@ -298,7 +297,7 @@ define <8 x double> @shuffle_v8f64_091b2d3f(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_09ab1def(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_09ab1def:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm0[1,0,2,2]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
@@ -312,7 +311,7 @@ define <8 x double> @shuffle_v8f64_00014445(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00014445:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -324,7 +323,7 @@ define <8 x double> @shuffle_v8f64_00204464(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00204464:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,2,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -336,7 +335,7 @@ define <8 x double> @shuffle_v8f64_03004744(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_03004744:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -348,7 +347,7 @@ define <8 x double> @shuffle_v8f64_10005444(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_10005444:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -360,7 +359,7 @@ define <8 x double> @shuffle_v8f64_22006644(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_22006644:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -372,7 +371,7 @@ define <8 x double> @shuffle_v8f64_33307774(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_33307774:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -384,7 +383,7 @@ define <8 x double> @shuffle_v8f64_32107654(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_32107654:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,2,1,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -396,7 +395,7 @@ define <8 x double> @shuffle_v8f64_00234467(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00234467:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -404,15 +403,15 @@ define <8 x double> @shuffle_v8f64_00234467(<8 x double> %a, <8 x double> %b) {
   ret <8 x double> %shuffle
 }
 
-define <8 x double> @shuffle_v8f64_00224466(<8 x double> %a, <8 x double> %b) {\r
-; ALL-LABEL: shuffle_v8f64_00224466:\r
-; ALL:       # BB#0:\r
-; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]\r
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 \r
-; ALL-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]\r
-; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0\r
-; ALL-NEXT:    retq\r
-  %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>\r
+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:    vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
+; 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>
   ret <8 x double> %shuffle
 }
 
@@ -420,7 +419,7 @@ define <8 x double> @shuffle_v8f64_10325476(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_10325476:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,0,3,2]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -432,7 +431,7 @@ define <8 x double> @shuffle_v8f64_11335577(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_11335577:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,1,3,3]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -444,7 +443,7 @@ define <8 x double> @shuffle_v8f64_10235467(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_10235467:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,0,2,3]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -456,7 +455,7 @@ define <8 x double> @shuffle_v8f64_10225466(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_10225466:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,0,2,2]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -468,7 +467,7 @@ define <8 x double> @shuffle_v8f64_00015444(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00015444:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,0,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -480,7 +479,7 @@ define <8 x double> @shuffle_v8f64_00204644(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00204644:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -492,7 +491,7 @@ define <8 x double> @shuffle_v8f64_03004474(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_03004474:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,0,3,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -504,7 +503,7 @@ define <8 x double> @shuffle_v8f64_10004444(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_10004444:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -516,7 +515,7 @@ define <8 x double> @shuffle_v8f64_22006446(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_22006446:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,0,0,2]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -528,7 +527,7 @@ define <8 x double> @shuffle_v8f64_33307474(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_33307474:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,0,3,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -540,7 +539,7 @@ define <8 x double> @shuffle_v8f64_32104567(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_32104567:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %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 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
@@ -551,7 +550,7 @@ define <8 x double> @shuffle_v8f64_00236744(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_00236744:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,3,0,0]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -559,13 +558,13 @@ define <8 x double> @shuffle_v8f64_00236744(<8 x double> %a, <8 x double> %b) {
   ret <8 x double> %shuffle
 }
 
-define <8 x double> @shuffle_v8f64_00226644(<8 x double> %a, <8 x double> %b) {\r
-; ALL-LABEL: shuffle_v8f64_00226644:\r
-; ALL:       # BB#0:\r
-; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]\r
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 \r
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,0,0]\r
-; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0\r
+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:    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
 ; 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 6, i32 6, i32 4, i32 4>
   ret <8 x double> %shuffle
@@ -575,7 +574,7 @@ define <8 x double> @shuffle_v8f64_10324567(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_10324567:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,0,3,2]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %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 3, i32 2, i32 4, i32 5, i32 6, i32 7>
@@ -586,7 +585,7 @@ define <8 x double> @shuffle_v8f64_11334567(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_11334567:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,1,3,3]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %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 1, i32 3, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -596,7 +595,7 @@ define <8 x double> @shuffle_v8f64_11334567(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_01235467(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_01235467:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -607,7 +606,7 @@ define <8 x double> @shuffle_v8f64_01235467(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_01235466(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_01235466:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -615,13 +614,13 @@ define <8 x double> @shuffle_v8f64_01235466(<8 x double> %a, <8 x double> %b) {
   ret <8 x double> %shuffle
 }
 
-define <8 x double> @shuffle_v8f64_002u6u44(<8 x double> %a, <8 x double> %b) {\r
-; ALL-LABEL: shuffle_v8f64_002u6u44:\r
-; ALL:       # BB#0:\r
-; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm0[0,0,2,2]\r
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 \r
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,1,0,0]\r
-; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0\r
+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:    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
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 undef, i32 6, i32 undef, i32 4, i32 4>
   ret <8 x double> %shuffle
@@ -631,7 +630,7 @@ 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:    vbroadcastsd %xmm0, %ymm1
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -643,7 +642,7 @@ define <8 x double> @shuffle_v8f64_103245uu(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_103245uu:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,0,3,2]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %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 3, i32 2, i32 4, i32 5, i32 undef, i32 undef>
@@ -654,7 +653,7 @@ define <8 x double> @shuffle_v8f64_1133uu67(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_1133uu67:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm0[1,1,3,3]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %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 1, i32 3, i32 3, i32 undef, i32 undef, i32 6, i32 7>
@@ -664,7 +663,7 @@ define <8 x double> @shuffle_v8f64_1133uu67(<8 x double> %a, <8 x double> %b) {
 define <8 x double> @shuffle_v8f64_0uu354uu(<8 x double> %a, <8 x double> %b) {
 ; ALL-LABEL: shuffle_v8f64_0uu354uu:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
 ; ALL-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -673,22 +672,22 @@ define <8 x double> @shuffle_v8f64_0uu354uu(<8 x double> %a, <8 x double> %b) {
 }
 
 define <8 x double> @shuffle_v8f64_uuu3uu66(<8 x double> %a, <8 x double> %b) {
-; ALL-LABEL: shuffle_v8f64_uuu3uu66:\r
-; ALL:       # BB#0:\r
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1 \r
-; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm1[0,0,2,2]\r
-; ALL-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0\r
-; ALL-NEXT:    retq\r
-  %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>\r
+; ALL-LABEL: shuffle_v8f64_uuu3uu66:
+; ALL:       # BB#0:
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
+; 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>
   ret <8 x double> %shuffle
 }
 
 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:    vextractf64x4 $1, %zmm0, %ymm2
 ; ALL-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[0,1],ymm2[0,1]
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3 
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm3
 ; 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]
@@ -704,18 +703,17 @@ define <8 x double> @shuffle_v8f64_c348cda0(<8 x double> %a, <8 x double> %b) {
 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:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm3 \r
-; ALL-NEXT:    vpermpd {{.*#+}} ymm4 = ymm3[0,1,1,3]\r
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3]\r
-; ALL-NEXT:    vmovddup {{.*#+}} ymm4 = ymm1[0,0,2,2]\r
-; ALL-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3]\r
-; ALL-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,1,1]\r
-; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2,3]\r
-; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm1 
+; ALL-NEXT:    vextractf64x4 $1, %zmm0, %ymm2
+; 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]
+; ALL-NEXT:    vextractf64x4 $1, %zmm1, %ymm1
 ; ALL-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[3,1,2,3]
 ; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
-; ALL-NEXT:    vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; ALL-NEXT:    vinsertf64x4 $1, %ymm3, %zmm0, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 15, i32 5, i32 1, i32 1, i32 2, i32 3, i32 5, i32 10>
   ret <8 x double> %shuffle
@@ -766,7 +764,7 @@ define <8 x i64> @shuffle_v8i64_00003000(<8 x i64> %a, <8 x i64> %b) {
 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:    vextracti64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
 ; ALL-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3,4,5],ymm1[6,7]
@@ -779,10 +777,10 @@ define <8 x i64> @shuffle_v8i64_00040000(<8 x i64> %a, <8 x i64> %b) {
 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:    vextracti64x4 $1, %zmm0, %ymm1
+; 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:    vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[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 5, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -792,10 +790,10 @@ define <8 x i64> @shuffle_v8i64_00500000(<8 x i64> %a, <8 x i64> %b) {
 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:    vextracti64x4 $1, %zmm0, %ymm1
+; 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:    vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
 ; 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 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -805,10 +803,10 @@ define <8 x i64> @shuffle_v8i64_06000000(<8 x i64> %a, <8 x i64> %b) {
 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:    vextracti64x4 $1, %zmm0, %ymm1
+; 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:    vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
@@ -818,7 +816,7 @@ 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:    vextracti64x4 $1, %zmm0, %ymm1 
+; 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
@@ -852,8 +850,8 @@ define <8 x i64> @shuffle_v8i64_00001111(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_81a3c5e7(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_81a3c5e7:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2 
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
@@ -877,9 +875,9 @@ 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, %zmm0, %ymm2 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
 ; ALL-NEXT:    vinserti128 $1, %xmm2, %ymm2, %ymm2
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3 
+; 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
@@ -894,8 +892,8 @@ define <8 x i64> @shuffle_v8i64_08084c4c(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_8823cc67(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_8823cc67:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2 
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
 ; ALL-NEXT:    vpbroadcastq %xmm3, %ymm3
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
 ; ALL-NEXT:    vpbroadcastq %xmm1, %ymm1
@@ -939,9 +937,9 @@ define <8 x i64> @shuffle_v8i64_9810dc54(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_08194c5d(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_08194c5d:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,0,2,1]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm3 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm3
 ; ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,1,1,3]
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,0,2,1]
@@ -956,9 +954,9 @@ define <8 x i64> @shuffle_v8i64_08194c5d(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_2a3b6e7f(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_2a3b6e7f:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm3 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm3
 ; ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[2,1,3,3]
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
@@ -988,12 +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
@@ -1002,7 +999,7 @@ define <8 x i64> @shuffle_v8i64_08991abb(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_091b2d3f(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_091b2d3f:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpermq {{.*#+}} ymm3 = ymm0[2,1,3,3]
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,1,1,3]
@@ -1016,7 +1013,7 @@ define <8 x i64> @shuffle_v8i64_091b2d3f(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_09ab1def(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_09ab1def:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2 
+; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm3 = ymm0[2,3,2,3,6,7,6,7]
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; ALL-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
@@ -1030,7 +1027,7 @@ define <8 x i64> @shuffle_v8i64_00014445(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00014445:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,0,1]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1042,7 +1039,7 @@ define <8 x i64> @shuffle_v8i64_00204464(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00204464:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,2,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1054,7 +1051,7 @@ define <8 x i64> @shuffle_v8i64_03004744(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_03004744:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,3,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1066,7 +1063,7 @@ define <8 x i64> @shuffle_v8i64_10005444(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_10005444:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,0,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1078,7 +1075,7 @@ define <8 x i64> @shuffle_v8i64_22006644(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_22006644:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,2,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1090,7 +1087,7 @@ define <8 x i64> @shuffle_v8i64_33307774(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_33307774:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,3,3,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1102,7 +1099,7 @@ define <8 x i64> @shuffle_v8i64_32107654(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_32107654:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,2,1,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1114,7 +1111,7 @@ define <8 x i64> @shuffle_v8i64_00234467(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00234467:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,2,3]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1126,7 +1123,7 @@ define <8 x i64> @shuffle_v8i64_00224466(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00224466:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[0,1,0,1,4,5,4,5]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1138,7 +1135,7 @@ define <8 x i64> @shuffle_v8i64_10325476(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_10325476:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1150,7 +1147,7 @@ define <8 x i64> @shuffle_v8i64_11335577(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_11335577:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1162,7 +1159,7 @@ define <8 x i64> @shuffle_v8i64_10235467(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_10235467:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[1,0,2,3]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,0,2,3]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1174,7 +1171,7 @@ define <8 x i64> @shuffle_v8i64_10225466(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_10225466:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[1,0,2,2]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,0,2,2]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1186,7 +1183,7 @@ define <8 x i64> @shuffle_v8i64_00015444(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00015444:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,0,1]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,0,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1198,7 +1195,7 @@ define <8 x i64> @shuffle_v8i64_00204644(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00204644:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,2,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1210,7 +1207,7 @@ define <8 x i64> @shuffle_v8i64_03004474(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_03004474:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,3,0,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,0,3,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1222,7 +1219,7 @@ define <8 x i64> @shuffle_v8i64_10004444(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_10004444:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[1,0,0,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1234,7 +1231,7 @@ define <8 x i64> @shuffle_v8i64_22006446(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_22006446:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,2,0,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,0,0,2]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1246,7 +1243,7 @@ define <8 x i64> @shuffle_v8i64_33307474(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_33307474:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[3,3,3,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[3,0,3,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1258,7 +1255,7 @@ define <8 x i64> @shuffle_v8i64_32104567(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_32104567:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[3,2,1,0]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %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 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
@@ -1269,7 +1266,7 @@ define <8 x i64> @shuffle_v8i64_00236744(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00236744:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[0,0,2,3]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1281,7 +1278,7 @@ define <8 x i64> @shuffle_v8i64_00226644(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_00226644:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[0,1,0,1,4,5,4,5]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,2,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1293,7 +1290,7 @@ define <8 x i64> @shuffle_v8i64_10324567(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_10324567:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %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 3, i32 2, i32 4, i32 5, i32 6, i32 7>
@@ -1304,7 +1301,7 @@ define <8 x i64> @shuffle_v8i64_11334567(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_11334567:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %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 1, i32 3, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -1314,7 +1311,7 @@ define <8 x i64> @shuffle_v8i64_11334567(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_01235467(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_01235467:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[1,0,2,3]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -1325,7 +1322,7 @@ define <8 x i64> @shuffle_v8i64_01235467(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_01235466(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_01235466:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[1,0,2,2]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -1337,7 +1334,7 @@ define <8 x i64> @shuffle_v8i64_002u6u44(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_002u6u44:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[0,1,0,1,4,5,4,5]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,1,0,0]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1349,7 +1346,7 @@ 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:    vpbroadcastq %xmm0, %ymm1
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,2,2,3]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; ALL-NEXT:    retq
@@ -1361,7 +1358,7 @@ define <8 x i64> @shuffle_v8i64_103245uu(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_103245uu:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[2,3,0,1,6,7,4,5]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %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 3, i32 2, i32 4, i32 5, i32 undef, i32 undef>
@@ -1372,7 +1369,7 @@ define <8 x i64> @shuffle_v8i64_1133uu67(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_1133uu67:
 ; ALL:       # BB#0:
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm0[2,3,2,3,6,7,6,7]
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %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 1, i32 3, i32 3, i32 undef, i32 undef, i32 6, i32 7>
@@ -1382,7 +1379,7 @@ define <8 x i64> @shuffle_v8i64_1133uu67(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_0uu354uu(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_0uu354uu:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -1393,7 +1390,7 @@ define <8 x i64> @shuffle_v8i64_0uu354uu(<8 x i64> %a, <8 x i64> %b) {
 define <8 x i64> @shuffle_v8i64_uuu3uu66(<8 x i64> %a, <8 x i64> %b) {
 ; ALL-LABEL: shuffle_v8i64_uuu3uu66:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
 ; ALL-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,4,5]
 ; ALL-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; ALL-NEXT:    retq
@@ -1404,9 +1401,9 @@ 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, %zmm0, %ymm0 
+; ALL-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
 ; ALL-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[0,1,0,1]
-; ALL-NEXT:    vextracti64x4 $1, %zmm1, %ymm3 
+; 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]