[llvm] faecc73 - [DAG] isSplatValue - node is a splat if all demanded elts have the same whole constant value (#74443)

via llvm-commits llvm-commits at lists.llvm.org
Fri Dec 8 02:53:56 PST 2023


Author: Simon Pilgrim
Date: 2023-12-08T10:53:51Z
New Revision: faecc736e2ac3cd8c77bebf41b1ed2e2d8cb575f

URL: https://github.com/llvm/llvm-project/commit/faecc736e2ac3cd8c77bebf41b1ed2e2d8cb575f
DIFF: https://github.com/llvm/llvm-project/commit/faecc736e2ac3cd8c77bebf41b1ed2e2d8cb575f.diff

LOG:  [DAG] isSplatValue - node is a splat if all demanded elts have the same whole constant value (#74443)

Added: 
    

Modified: 
    llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
    llvm/test/CodeGen/ARM/vector-store.ll
    llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-shuffles.ll
    llvm/test/CodeGen/X86/fold-pcmpeqd-2.ll
    llvm/test/CodeGen/X86/var-permute-256.ll
    llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
    llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
    llvm/test/CodeGen/X86/vector-shift-ashr-128.ll
    llvm/test/CodeGen/X86/vector-shift-ashr-256.ll
    llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
    llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
    llvm/test/CodeGen/X86/vector-shift-shl-128.ll
    llvm/test/CodeGen/X86/vector-shift-shl-256.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 51ae8b703e50f2..5be1892a44f6dd 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -2881,6 +2881,12 @@ bool SelectionDAG::isSplatValue(SDValue V, const APInt &DemandedElts,
   }
   }
 
+  // Fallback - this is a splat if all demanded elts are the same constant.
+  if (computeKnownBits(V, DemandedElts, Depth).isConstant()) {
+    UndefElts = ~DemandedElts;
+    return true;
+  }
+
   return false;
 }
 

diff  --git a/llvm/test/CodeGen/ARM/vector-store.ll b/llvm/test/CodeGen/ARM/vector-store.ll
index a8a1031637afc4..9c8ea7a2c440c8 100644
--- a/llvm/test/CodeGen/ARM/vector-store.ll
+++ b/llvm/test/CodeGen/ARM/vector-store.ll
@@ -403,17 +403,14 @@ define void @v3i8store(ptr %p) {
 ; CHECK-LABEL: v3i8store:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    vmov.i32 d16, #0xff
-; CHECK-NEXT:    mov r1, sp
-; CHECK-NEXT:    vmov.i32 d17, #0x0
-; CHECK-NEXT:    movs r2, #0
-; CHECK-NEXT:    vand d16, d17, d16
-; CHECK-NEXT:    vst1.32 {d16[0]}, [r1:32]
-; CHECK-NEXT:    vld1.32 {d16[0]}, [r1:32]
+; CHECK-NEXT:    movs r1, #0
+; CHECK-NEXT:    mov r2, sp
+; CHECK-NEXT:    str r1, [sp]
+; CHECK-NEXT:    vld1.32 {d16[0]}, [r2:32]
+; CHECK-NEXT:    strb r1, [r0, #2]
 ; CHECK-NEXT:    vmovl.u16 q8, d16
-; CHECK-NEXT:    strb r2, [r0, #2]
-; CHECK-NEXT:    vmov.32 r1, d16[0]
-; CHECK-NEXT:    strh r1, [r0]
+; CHECK-NEXT:    vmov.32 r2, d16[0]
+; CHECK-NEXT:    strh r2, [r0]
 ; CHECK-NEXT:    add sp, #4
 ; CHECK-NEXT:    bx lr
   store <3 x i8> zeroinitializer, ptr %p, align 4

diff  --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-shuffles.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-shuffles.ll
index a56a81f5f793bc..769bfe8cd5ba97 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-shuffles.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-shuffles.ll
@@ -244,14 +244,14 @@ define <8 x i64> @vrgather_shuffle_vx_v8i64(<8 x i64> %x) {
 ; RV32-NEXT:    addi a0, a0, %lo(.LCPI13_0)
 ; RV32-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
 ; RV32-NEXT:    vle16.v v16, (a0)
-; RV32-NEXT:    vrgatherei16.vv v12, v8, v16
+; RV32-NEXT:    vmv.v.i v20, 5
 ; RV32-NEXT:    lui a0, %hi(.LCPI13_1)
 ; RV32-NEXT:    addi a0, a0, %lo(.LCPI13_1)
-; RV32-NEXT:    vle16.v v8, (a0)
-; RV32-NEXT:    li a0, 140
+; RV32-NEXT:    vle16.v v17, (a0)
+; RV32-NEXT:    li a0, 115
 ; RV32-NEXT:    vmv.s.x v0, a0
-; RV32-NEXT:    vmv.v.i v16, 5
-; RV32-NEXT:    vrgatherei16.vv v12, v16, v8, v0.t
+; RV32-NEXT:    vrgatherei16.vv v12, v20, v16
+; RV32-NEXT:    vrgatherei16.vv v12, v8, v17, v0.t
 ; RV32-NEXT:    vmv.v.v v8, v12
 ; RV32-NEXT:    ret
 ;

diff  --git a/llvm/test/CodeGen/X86/fold-pcmpeqd-2.ll b/llvm/test/CodeGen/X86/fold-pcmpeqd-2.ll
index 5dcb1d63207d14..493e7aa7039e3f 100644
--- a/llvm/test/CodeGen/X86/fold-pcmpeqd-2.ll
+++ b/llvm/test/CodeGen/X86/fold-pcmpeqd-2.ll
@@ -51,11 +51,6 @@ define void @program_1(ptr %dest, ptr %t0, <4 x float> %p0, <4 x float> %p1, <4
 ; X86-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
 ; X86-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
 ; X86-NEXT:    movaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
-; X86-NEXT:    xorps %xmm0, %xmm0
-; X86-NEXT:    movaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
-; X86-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
-; X86-NEXT:    mulps %xmm0, %xmm0
-; X86-NEXT:    movaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
 ; X86-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
 ; X86-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
 ; X86-NEXT:    movaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
@@ -64,8 +59,10 @@ define void @program_1(ptr %dest, ptr %t0, <4 x float> %p0, <4 x float> %p1, <4
 ; X86-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
 ; X86-NEXT:    cmpunordps %xmm0, %xmm0
 ; X86-NEXT:    movaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
+; X86-NEXT:    xorps %xmm0, %xmm0
+; X86-NEXT:    movaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
 ; X86-NEXT:    movaps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 ## 16-byte Reload
-; X86-NEXT:    minps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-NEXT:    minps %xmm0, %xmm0
 ; X86-NEXT:    movaps %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) ## 16-byte Spill
 ; X86-NEXT:    xorps %xmm0, %xmm0
 ; X86-NEXT:    movaps %xmm0, {{[0-9]+}}(%esp)
@@ -135,11 +132,6 @@ define void @program_1(ptr %dest, ptr %t0, <4 x float> %p0, <4 x float> %p1, <4
 ; X64-NEXT:    movaps (%rsp), %xmm0 ## 16-byte Reload
 ; X64-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; X64-NEXT:    movaps %xmm0, (%rsp) ## 16-byte Spill
-; X64-NEXT:    xorps %xmm0, %xmm0
-; X64-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
-; X64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
-; X64-NEXT:    mulps %xmm0, %xmm0
-; X64-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
 ; X64-NEXT:    movaps (%rsp), %xmm0 ## 16-byte Reload
 ; X64-NEXT:    mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; X64-NEXT:    movaps %xmm0, (%rsp) ## 16-byte Spill
@@ -148,8 +140,10 @@ define void @program_1(ptr %dest, ptr %t0, <4 x float> %p0, <4 x float> %p1, <4
 ; X64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
 ; X64-NEXT:    cmpunordps %xmm0, %xmm0
 ; X64-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
+; X64-NEXT:    xorps %xmm0, %xmm0
+; X64-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
 ; X64-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 16-byte Reload
-; X64-NEXT:    minps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-NEXT:    minps %xmm0, %xmm0
 ; X64-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) ## 16-byte Spill
 ; X64-NEXT:    xorl %ebx, %ebx
 ; X64-NEXT:    xorps %xmm3, %xmm3

diff  --git a/llvm/test/CodeGen/X86/var-permute-256.ll b/llvm/test/CodeGen/X86/var-permute-256.ll
index 6c07c4ca523f8f..853d8a278960d7 100644
--- a/llvm/test/CodeGen/X86/var-permute-256.ll
+++ b/llvm/test/CodeGen/X86/var-permute-256.ll
@@ -25,18 +25,20 @@ define <4 x i64> @var_shuffle_v4i64(<4 x i64> %v, <4 x i64> %indices) nounwind {
 ;
 ; AVX1-LABEL: var_shuffle_v4i64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpaddq %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm3 = [2,2]
+; AVX1-NEXT:    # xmm3 = mem[0,0]
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm4
 ; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm4
-; AVX1-NEXT:    vpermilpd %ymm4, %ymm2, %ymm2
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm4, %ymm2
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilpd %ymm4, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vblendvpd %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vblendvpd %ymm3, %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: var_shuffle_v4i64:
@@ -88,15 +90,16 @@ define <8 x i32> @var_shuffle_v8i32(<8 x i32> %v, <8 x i32> %indices) nounwind {
 ;
 ; AVX1-LABEL: var_shuffle_v8i32:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpermilps %ymm1, %ymm2, %ymm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vbroadcastss {{.*#+}} xmm3 = [3,3,3,3]
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,2,3]
+; AVX1-NEXT:    vpermilps %ymm1, %ymm3, %ymm3
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vblendvps %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vblendvps %ymm2, %ymm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; INT256-LABEL: var_shuffle_v8i32:
@@ -445,18 +448,20 @@ define <4 x double> @var_shuffle_v4f64(<4 x double> %v, <4 x i64> %indices) noun
 ;
 ; AVX1-LABEL: var_shuffle_v4f64:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpaddq %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm3 = [2,2]
+; AVX1-NEXT:    # xmm3 = mem[0,0]
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm4
 ; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm4
-; AVX1-NEXT:    vpermilpd %ymm4, %ymm2, %ymm2
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm4, %ymm2
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilpd %ymm4, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vblendvpd %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vblendvpd %ymm3, %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: var_shuffle_v4f64:
@@ -508,15 +513,16 @@ define <8 x float> @var_shuffle_v8f32(<8 x float> %v, <8 x i32> %indices) nounwi
 ;
 ; AVX1-LABEL: var_shuffle_v8f32:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,2,3]
-; AVX1-NEXT:    vpermilps %ymm1, %ymm2, %ymm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vbroadcastss {{.*#+}} xmm3 = [3,3,3,3]
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,2,3]
+; AVX1-NEXT:    vpermilps %ymm1, %ymm3, %ymm3
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vblendvps %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vblendvps %ymm2, %ymm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; INT256-LABEL: var_shuffle_v8f32:
@@ -569,17 +575,19 @@ define <4 x i64> @var_shuffle_v4i64_from_v2i64(<2 x i64> %v, <4 x i64> %indices)
 ; AVX1-LABEL: var_shuffle_v4i64_from_v2i64:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpaddq %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm3 = [2,2]
+; AVX1-NEXT:    # xmm3 = mem[0,0]
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm4
 ; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm3
-; AVX1-NEXT:    vpermilpd %ymm3, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-NEXT:    vpermilpd %ymm3, %ymm0, %ymm2
-; AVX1-NEXT:    vblendvpd %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm1
+; AVX1-NEXT:    vblendvpd %ymm3, %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: var_shuffle_v4i64_from_v2i64:
@@ -633,14 +641,15 @@ define <8 x i32> @var_shuffle_v8i32_from_v4i32(<4 x i32> %v, <8 x i32> %indices)
 ; AVX1-LABEL: var_shuffle_v8i32_from_v4i32:
 ; AVX1:       # %bb.0: # %entry
 ; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vbroadcastss {{.*#+}} xmm3 = [3,3,3,3]
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm3
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vblendvps %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vblendvps %ymm2, %ymm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; INT256-LABEL: var_shuffle_v8i32_from_v4i32:
@@ -990,17 +999,19 @@ define <4 x double> @var_shuffle_v4f64_from_v2f64(<2 x double> %v, <4 x i64> %in
 ; AVX1-LABEL: var_shuffle_v4f64_from_v2f64:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpaddq %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm3 = [2,2]
+; AVX1-NEXT:    # xmm3 = mem[0,0]
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm4
 ; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm3
-; AVX1-NEXT:    vpermilpd %ymm3, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-NEXT:    vpermilpd %ymm3, %ymm0, %ymm2
-; AVX1-NEXT:    vblendvpd %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm1
+; AVX1-NEXT:    vblendvpd %ymm3, %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: var_shuffle_v4f64_from_v2f64:
@@ -1054,14 +1065,15 @@ define <8 x float> @var_shuffle_v8f32_from_v4f32(<4 x float> %v, <8 x i32> %indi
 ; AVX1-LABEL: var_shuffle_v8f32_from_v4f32:
 ; AVX1:       # %bb.0: # %entry
 ; AVX1-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vbroadcastss {{.*#+}} xmm3 = [3,3,3,3]
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm3
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vblendvps %ymm1, %ymm2, %ymm0, %ymm0
+; AVX1-NEXT:    vblendvps %ymm2, %ymm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; INT256-LABEL: var_shuffle_v8f32_from_v4f32:
@@ -1271,20 +1283,22 @@ define <4 x i64> @var_shuffle_v4i64_with_v16i8_indices(<4 x i64> %v, <16 x i8> %
 ;
 ; AVX1-LABEL: var_shuffle_v4i64_with_v16i8_indices:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpsrld $16, %xmm1, %xmm2
-; AVX1-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,zero,zero,zero,zero,xmm2[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpmovzxbq {{.*#+}} xmm2 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpsrld $16, %xmm1, %xmm1
 ; AVX1-NEXT:    vpmovzxbq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,zero,zero,zero,zero,xmm1[1],zero,zero,zero,zero,zero,zero,zero
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,2,3]
 ; AVX1-NEXT:    vpaddq %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vmovddup {{.*#+}} xmm3 = [2,2]
+; AVX1-NEXT:    # xmm3 = mem[0,0]
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm4
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm4
-; AVX1-NEXT:    vpermilpd %ymm4, %ymm3, %ymm3
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm4, %ymm2
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vpermilpd %ymm4, %ymm0, %ymm0
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpgtq {{\.?LCPI[0-9]+_[0-9]+}}+16(%rip), %xmm2, %xmm2
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX1-NEXT:    vblendvpd %ymm1, %ymm3, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vblendvpd %ymm3, %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: var_shuffle_v4i64_with_v16i8_indices:

diff  --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
index c9cbb0994810f3..e2195f1fc25a13 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
@@ -156,64 +156,62 @@ define void @load_i8_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
 ; SSE-LABEL: load_i8_stride4_vf8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa (%rdi), %xmm1
+; SSE-NEXT:    movdqa (%rdi), %xmm3
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm4
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,0,255,0,255,0,255,0]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm4[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    packuswb %xmm4, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pxor %xmm7, %xmm7
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3],xmm3[4],xmm7[4],xmm3[5],xmm7[5],xmm3[6],xmm7[6],xmm3[7],xmm7[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm7[8],xmm5[9],xmm7[9],xmm5[10],xmm7[10],xmm5[11],xmm7[11],xmm5[12],xmm7[12],xmm5[13],xmm7[13],xmm5[14],xmm7[14],xmm5[15],xmm7[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm6[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
-; SSE-NEXT:    packuswb %xmm8, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm8, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT:    pxor %xmm6, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3],xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1]
+; SSE-NEXT:    packuswb %xmm7, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,3,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm7, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm5, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
 ; SSE-NEXT:    movq %xmm0, (%rsi)
-; SSE-NEXT:    movq %xmm7, (%rdx)
-; SSE-NEXT:    movq %xmm1, (%rcx)
-; SSE-NEXT:    movq %xmm2, (%r8)
+; SSE-NEXT:    movq %xmm6, (%rdx)
+; SSE-NEXT:    movq %xmm5, (%rcx)
+; SSE-NEXT:    movq %xmm1, (%r8)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: load_i8_stride4_vf8:
@@ -296,111 +294,111 @@ define void @load_i8_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-LABEL: load_i8_stride4_vf16:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movdqa (%rdi), %xmm1
-; SSE-NEXT:    movdqa 16(%rdi), %xmm2
-; SSE-NEXT:    movdqa 32(%rdi), %xmm8
-; SSE-NEXT:    movdqa 48(%rdi), %xmm13
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,0,255,0,255,0,255,0]
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    packuswb %xmm4, %xmm0
-; SSE-NEXT:    pxor %xmm10, %xmm10
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm10[8],xmm3[9],xmm10[9],xmm3[10],xmm10[10],xmm3[11],xmm10[11],xmm3[12],xmm10[12],xmm3[13],xmm10[13],xmm3[14],xmm10[14],xmm3[15],xmm10[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3],xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm6[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
-; SSE-NEXT:    movdqa %xmm8, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm10[8],xmm5[9],xmm10[9],xmm5[10],xmm10[10],xmm5[11],xmm10[11],xmm5[12],xmm10[12],xmm5[13],xmm10[13],xmm5[14],xmm10[14],xmm5[15],xmm10[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm6[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm8, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm10[0],xmm6[1],xmm10[1],xmm6[2],xmm10[2],xmm6[3],xmm10[3],xmm6[4],xmm10[4],xmm6[5],xmm10[5],xmm6[6],xmm10[6],xmm6[7],xmm10[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm11[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
-; SSE-NEXT:    packuswb %xmm7, %xmm14
-; SSE-NEXT:    movdqa %xmm2, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm10[8],xmm7[9],xmm10[9],xmm7[10],xmm10[10],xmm7[11],xmm10[11],xmm7[12],xmm10[12],xmm7[13],xmm10[13],xmm7[14],xmm10[14],xmm7[15],xmm10[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm9[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm2, %xmm9
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3],xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm12[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm11[0],xmm15[1],xmm11[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm11
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm10[8],xmm11[9],xmm10[9],xmm11[10],xmm10[10],xmm11[11],xmm10[11],xmm11[12],xmm10[12],xmm11[13],xmm10[13],xmm11[14],xmm10[14],xmm11[15],xmm10[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3],xmm12[4],xmm10[4],xmm12[5],xmm10[5],xmm12[6],xmm10[6],xmm12[7],xmm10[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm12[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT:    packuswb %xmm15, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm14[0,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm1, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm13, %xmm8
+; SSE-NEXT:    movdqa 16(%rdi), %xmm3
+; SSE-NEXT:    movdqa 32(%rdi), %xmm9
+; SSE-NEXT:    movdqa 48(%rdi), %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    pand %xmm2, %xmm7
+; SSE-NEXT:    movdqa %xmm9, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm6
+; SSE-NEXT:    movdqa %xmm9, %xmm11
+; SSE-NEXT:    pand %xmm2, %xmm11
+; SSE-NEXT:    packuswb %xmm7, %xmm11
+; SSE-NEXT:    movdqa %xmm3, %xmm7
+; SSE-NEXT:    movdqa %xmm3, %xmm8
+; SSE-NEXT:    movdqa %xmm3, %xmm12
+; SSE-NEXT:    pand %xmm2, %xmm12
 ; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pand %xmm1, %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm12, %xmm2
+; SSE-NEXT:    packuswb %xmm11, %xmm2
+; SSE-NEXT:    pxor %xmm11, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm11[8],xmm0[9],xmm11[9],xmm0[10],xmm11[10],xmm0[11],xmm11[11],xmm0[12],xmm11[12],xmm0[13],xmm11[13],xmm0[14],xmm11[14],xmm0[15],xmm11[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm11[0],xmm4[1],xmm11[1],xmm4[2],xmm11[2],xmm4[3],xmm11[3],xmm4[4],xmm11[4],xmm4[5],xmm11[5],xmm4[6],xmm11[6],xmm4[7],xmm11[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm11[8],xmm5[9],xmm11[9],xmm5[10],xmm11[10],xmm5[11],xmm11[11],xmm5[12],xmm11[12],xmm5[13],xmm11[13],xmm5[14],xmm11[14],xmm5[15],xmm11[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm5[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm11[0],xmm6[1],xmm11[1],xmm6[2],xmm11[2],xmm6[3],xmm11[3],xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm12[0],xmm14[1],xmm12[1]
+; SSE-NEXT:    packuswb %xmm13, %xmm14
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm11[8],xmm7[9],xmm11[9],xmm7[10],xmm11[10],xmm7[11],xmm11[11],xmm7[12],xmm11[12],xmm7[13],xmm11[13],xmm7[14],xmm11[14],xmm7[15],xmm11[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm11[0],xmm8[1],xmm11[1],xmm8[2],xmm11[2],xmm8[3],xmm11[3],xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm13[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm12[0],xmm15[1],xmm12[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    movdqa %xmm1, %xmm13
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm11[8],xmm13[9],xmm11[9],xmm13[10],xmm11[10],xmm13[11],xmm11[11],xmm13[12],xmm11[12],xmm13[13],xmm11[13],xmm13[14],xmm11[14],xmm13[15],xmm11[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3],xmm12[4],xmm11[4],xmm12[5],xmm11[5],xmm12[6],xmm11[6],xmm12[7],xmm11[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm12[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm15, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,3],xmm14[0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm8[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm9[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
 ; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
 ; SSE-NEXT:    packuswb %xmm3, %xmm5
 ; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm4[0,3]
-; SSE-NEXT:    movdqa %xmm0, (%rsi)
-; SSE-NEXT:    movaps %xmm10, (%rdx)
+; SSE-NEXT:    movdqa %xmm2, (%rsi)
+; SSE-NEXT:    movaps %xmm11, (%rdx)
 ; SSE-NEXT:    movaps %xmm1, (%rcx)
 ; SSE-NEXT:    movaps %xmm5, (%r8)
 ; SSE-NEXT:    retq
@@ -530,249 +528,247 @@ define void @load_i8_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $136, %rsp
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdi), %xmm5
-; SSE-NEXT:    movdqa 96(%rdi), %xmm15
-; SSE-NEXT:    movdqa 112(%rdi), %xmm14
-; SSE-NEXT:    movdqa (%rdi), %xmm12
-; SSE-NEXT:    movdqa 16(%rdi), %xmm11
-; SSE-NEXT:    movdqa 32(%rdi), %xmm7
-; SSE-NEXT:    movdqa 48(%rdi), %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,0,255,0,255,0,255,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm2
-; SSE-NEXT:    packuswb %xmm0, %xmm2
-; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa 80(%rdi), %xmm9
+; SSE-NEXT:    movdqa 96(%rdi), %xmm11
+; SSE-NEXT:    movdqa 112(%rdi), %xmm12
+; SSE-NEXT:    movdqa (%rdi), %xmm13
+; SSE-NEXT:    movdqa 16(%rdi), %xmm14
+; SSE-NEXT:    movdqa 32(%rdi), %xmm2
+; SSE-NEXT:    movdqa 48(%rdi), %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,0,255,0,255,0,255,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm15
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    packuswb %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm7
+; SSE-NEXT:    pand %xmm8, %xmm7
+; SSE-NEXT:    packuswb %xmm0, %xmm7
+; SSE-NEXT:    packuswb %xmm1, %xmm7
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    packuswb %xmm0, %xmm6
-; SSE-NEXT:    packuswb %xmm1, %xmm6
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    pand %xmm4, %xmm8
+; SSE-NEXT:    packuswb %xmm0, %xmm8
+; SSE-NEXT:    packuswb %xmm1, %xmm8
+; SSE-NEXT:    pxor %xmm2, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm5, %xmm8
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
+; SSE-NEXT:    packuswb %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm5[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm13, %xmm14
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
 ; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm9, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,3],xmm8[0,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1],xmm14[2],xmm2[2],xmm14[3],xmm2[3],xmm14[4],xmm2[4],xmm14[5],xmm2[5],xmm14[6],xmm2[6],xmm14[7],xmm2[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm14[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm5[0,3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm15
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
-; SSE-NEXT:    packuswb %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm10
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm9
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
-; SSE-NEXT:    movdqa %xmm5, %xmm8
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3],xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm2[0],xmm15[1],xmm2[1],xmm15[2],xmm2[2],xmm15[3],xmm2[3],xmm15[4],xmm2[4],xmm15[5],xmm2[5],xmm15[6],xmm2[6],xmm15[7],xmm2[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm13
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm2[8],xmm13[9],xmm2[9],xmm13[10],xmm2[10],xmm13[11],xmm2[11],xmm13[12],xmm2[12],xmm13[13],xmm2[13],xmm13[14],xmm2[14],xmm13[15],xmm2[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1],xmm12[2],xmm2[2],xmm12[3],xmm2[3],xmm12[4],xmm2[4],xmm12[5],xmm2[5],xmm12[6],xmm2[6],xmm12[7],xmm2[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm2[8],xmm11[9],xmm2[9],xmm11[10],xmm2[10],xmm11[11],xmm2[11],xmm11[12],xmm2[12],xmm11[13],xmm2[13],xmm11[14],xmm2[14],xmm11[15],xmm2[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3],xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[1,3,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
 ; SSE-NEXT:    packuswb %xmm1, %xmm4
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm3, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm1[0,3]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm7[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm0, %xmm11
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm2[0,3]
-; SSE-NEXT:    pand %xmm0, %xmm14
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm15
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm15[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm3[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm7
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    pand %xmm0, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm2[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm7[0,3]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm7
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm2[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshufd $231, (%rsp), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm7[0,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1]
-; SSE-NEXT:    packuswb %xmm7, %xmm11
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd $231, (%rsp), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm3[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
-; SSE-NEXT:    packuswb %xmm7, %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm11[0,3]
-; SSE-NEXT:    movdqa %xmm6, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rsi)
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
+; SSE-NEXT:    movdqa %xmm8, 16(%rsi)
+; SSE-NEXT:    movdqa %xmm7, (%rsi)
 ; SSE-NEXT:    movaps %xmm4, 16(%rdx)
-; SSE-NEXT:    movaps %xmm13, (%rdx)
-; SSE-NEXT:    movaps %xmm3, 16(%rcx)
-; SSE-NEXT:    movaps %xmm1, (%rcx)
-; SSE-NEXT:    movaps %xmm8, 16(%r8)
-; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps %xmm10, (%rcx)
+; SSE-NEXT:    movaps %xmm5, 16(%r8)
+; SSE-NEXT:    movaps %xmm14, (%r8)
 ; SSE-NEXT:    addq $136, %rsp
 ; SSE-NEXT:    retq
 ;
@@ -1023,522 +1019,518 @@ define void @load_i8_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
 ; SSE-LABEL: load_i8_stride4_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $664, %rsp # imm = 0x298
+; SSE-NEXT:    subq $600, %rsp # imm = 0x258
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm8
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rdi), %xmm6
+; SSE-NEXT:    movdqa 32(%rdi), %xmm15
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm14
-; SSE-NEXT:    movdqa 128(%rdi), %xmm15
-; SSE-NEXT:    movdqa 144(%rdi), %xmm10
-; SSE-NEXT:    movdqa 160(%rdi), %xmm11
-; SSE-NEXT:    movdqa 176(%rdi), %xmm3
-; SSE-NEXT:    movdqa 64(%rdi), %xmm13
-; SSE-NEXT:    movdqa 80(%rdi), %xmm7
+; SSE-NEXT:    movdqa 128(%rdi), %xmm4
+; SSE-NEXT:    movdqa 144(%rdi), %xmm7
+; SSE-NEXT:    movdqa 160(%rdi), %xmm10
+; SSE-NEXT:    movdqa 176(%rdi), %xmm12
+; SSE-NEXT:    movdqa 64(%rdi), %xmm6
+; SSE-NEXT:    movdqa 80(%rdi), %xmm13
 ; SSE-NEXT:    movdqa 96(%rdi), %xmm2
 ; SSE-NEXT:    movdqa 112(%rdi), %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,255,0,255,0,255,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,0,255,0,255,0,255,0]
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm9
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    packuswb %xmm0, %xmm2
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    packuswb %xmm0, %xmm2
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    movdqa (%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa (%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    packuswb %xmm0, %xmm2
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 240(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    movdqa 240(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    movdqa 224(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
 ; SSE-NEXT:    movdqa 208(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm2, %xmm5
-; SSE-NEXT:    packuswb %xmm0, %xmm5
-; SSE-NEXT:    packuswb %xmm1, %xmm5
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    pand %xmm2, %xmm9
+; SSE-NEXT:    packuswb %xmm0, %xmm9
+; SSE-NEXT:    packuswb %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    pxor %xmm8, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    movdqa %xmm13, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[1,3,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm9[0,3]
+; SSE-NEXT:    packuswb %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm5[0,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3],xmm11[4],xmm2[4],xmm11[5],xmm2[5],xmm11[6],xmm2[6],xmm11[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm6
+; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1],xmm10[2],xmm2[2],xmm10[3],xmm2[3],xmm10[4],xmm2[4],xmm10[5],xmm2[5],xmm10[6],xmm2[6],xmm10[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm9[0,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm6[0,3]
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1],xmm14[2],xmm8[2],xmm14[3],xmm8[3],xmm14[4],xmm8[4],xmm14[5],xmm8[5],xmm14[6],xmm8[6],xmm14[7],xmm8[7]
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm4[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3],xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm4[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm3[0],xmm15[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm1, %xmm15
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,3],xmm0[0,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm0[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    movdqa %xmm15, %xmm14
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm8[0],xmm12[1],xmm8[1],xmm12[2],xmm8[2],xmm12[3],xmm8[3],xmm12[4],xmm8[4],xmm12[5],xmm8[5],xmm12[6],xmm8[6],xmm12[7],xmm8[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm12[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm13
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm8[8],xmm13[9],xmm8[9],xmm13[10],xmm8[10],xmm13[11],xmm8[11],xmm13[12],xmm8[12],xmm13[13],xmm8[13],xmm13[14],xmm8[14],xmm13[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3],xmm15[4],xmm8[4],xmm15[5],xmm8[5],xmm15[6],xmm8[6],xmm15[7],xmm8[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm15[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm11, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm8[8],xmm7[9],xmm8[9],xmm7[10],xmm8[10],xmm7[11],xmm8[11],xmm7[12],xmm8[12],xmm7[13],xmm8[13],xmm7[14],xmm8[14],xmm7[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm11, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3],xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm1, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm10, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
 ; SSE-NEXT:    packuswb %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3],xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm1[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    packuswb %xmm0, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm3[0,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm0, %xmm14
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm4[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,3],xmm3[0,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm11[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm3[0,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm8, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm4[0,3]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm3[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm8[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,3],xmm4[0,3]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    pshuflw $231, (%rsp), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm4[0,3]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pand %xmm0, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm10[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm6[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm7
-; SSE-NEXT:    pand %xmm0, %xmm11
-; SSE-NEXT:    pand %xmm0, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm8[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm4[0,3]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm4[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,3],xmm7[0,3]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm4[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm7
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm12[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm15[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm7[0,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,3],xmm8[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
-; SSE-NEXT:    packuswb %xmm7, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm8
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm7[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
-; SSE-NEXT:    packuswb %xmm10, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm7
 ; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm8[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
-; SSE-NEXT:    packuswb %xmm8, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm12
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm8[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
-; SSE-NEXT:    packuswb %xmm11, %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm10[0,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm12[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
-; SSE-NEXT:    packuswb %xmm10, %xmm11
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm12
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm13 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm4[0],xmm13[1],xmm4[1]
-; SSE-NEXT:    packuswb %xmm10, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,3],xmm11[0,3]
-; SSE-NEXT:    movdqa %xmm5, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 16(%rsi)
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm3[0],xmm13[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm13
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,3],xmm12[0,3]
+; SSE-NEXT:    movdqa %xmm9, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 16(%rsi)
 ; SSE-NEXT:    movaps %xmm1, 48(%rdx)
-; SSE-NEXT:    movaps %xmm15, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rdx)
+; SSE-NEXT:    movaps %xmm5, (%rdx)
+; SSE-NEXT:    movaps %xmm10, 32(%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, 16(%rdx)
-; SSE-NEXT:    movaps %xmm6, 48(%rcx)
-; SSE-NEXT:    movaps %xmm3, 32(%rcx)
-; SSE-NEXT:    movaps %xmm2, 16(%rcx)
-; SSE-NEXT:    movaps %xmm12, (%rcx)
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps %xmm14, 32(%rcx)
+; SSE-NEXT:    movaps %xmm11, 16(%rcx)
+; SSE-NEXT:    movaps %xmm2, (%rcx)
 ; SSE-NEXT:    movaps %xmm13, 48(%r8)
 ; SSE-NEXT:    movaps %xmm8, 32(%r8)
 ; SSE-NEXT:    movaps %xmm7, 16(%r8)
-; SSE-NEXT:    movaps %xmm0, (%r8)
-; SSE-NEXT:    addq $664, %rsp # imm = 0x298
+; SSE-NEXT:    movaps %xmm6, (%r8)
+; SSE-NEXT:    addq $600, %rsp # imm = 0x258
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: load_i8_stride4_vf64:

diff  --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
index a2ad944c66f42e..be34bbd2540222 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
@@ -104,9 +104,9 @@ define void @load_i8_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [16711935,16711935,16711935,16711935]
+; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
@@ -130,10 +130,9 @@ define void @load_i8_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm9[2],xmm7[3],xmm9[3]
 ; SSE-NEXT:    packuswb %xmm7, %xmm7
 ; SSE-NEXT:    por %xmm7, %xmm4
-; SSE-NEXT:    movaps %xmm6, %xmm7
-; SSE-NEXT:    andps %xmm2, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm6[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm7
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm7, %xmm7
@@ -145,8 +144,8 @@ define void @load_i8_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
 ; SSE-NEXT:    packuswb %xmm6, %xmm6
 ; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    pand %xmm8, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,1,2,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[3,1,2,0]
+; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
@@ -225,10 +224,9 @@ define void @load_i8_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm3, %xmm8
 ; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [16711935,16711935,16711935,16711935]
 ; SSE-NEXT:    pand %xmm5, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
@@ -274,10 +272,9 @@ define void @load_i8_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm4, %xmm12
 ; SSE-NEXT:    pand %xmm11, %xmm12
 ; SSE-NEXT:    por %xmm9, %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm9
-; SSE-NEXT:    pand %xmm5, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm12[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm5, %xmm9
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm9[0,1,2,3,5,5,5,5]
@@ -318,8 +315,8 @@ define void @load_i8_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm11, %xmm3
 ; SSE-NEXT:    pandn %xmm4, %xmm11
 ; SSE-NEXT:    por %xmm3, %xmm11
-; SSE-NEXT:    pand %xmm11, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[3,1,2,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[3,1,2,0]
+; SSE-NEXT:    pand %xmm5, %xmm3
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[2,1,0,3,4,5,6,7]
@@ -470,79 +467,81 @@ define void @load_i8_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i8_stride6_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa 64(%rdi), %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 64(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rdi), %xmm5
-; SSE-NEXT:    movdqa 16(%rdi), %xmm2
-; SSE-NEXT:    movdqa 32(%rdi), %xmm8
+; SSE-NEXT:    movdqa 16(%rdi), %xmm1
+; SSE-NEXT:    movdqa 32(%rdi), %xmm7
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm6
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    pandn %xmm8, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm7, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    por %xmm0, %xmm6
 ; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm8, %xmm9
 ; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm10
-; SSE-NEXT:    pand %xmm4, %xmm11
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    pandn %xmm1, %xmm11
+; SSE-NEXT:    pand %xmm4, %xmm10
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm5, %xmm14
 ; SSE-NEXT:    pand %xmm4, %xmm14
-; SSE-NEXT:    movdqa 80(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm4, %xmm13
-; SSE-NEXT:    movdqa %xmm8, %xmm15
-; SSE-NEXT:    pand %xmm4, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm15
+; SSE-NEXT:    pand %xmm4, %xmm7
 ; SSE-NEXT:    pand %xmm4, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm12
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    pandn %xmm5, %xmm4
 ; SSE-NEXT:    pand %xmm0, %xmm5
-; SSE-NEXT:    por %xmm10, %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm10
-; SSE-NEXT:    pand %xmm7, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    por %xmm11, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm5[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    por %xmm9, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm12
-; SSE-NEXT:    por %xmm12, %xmm11
-; SSE-NEXT:    movdqa %xmm11, %xmm9
-; SSE-NEXT:    pand %xmm7, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[3,1,2,0]
+; SSE-NEXT:    pandn %xmm3, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm10[3,1,2,0]
+; SSE-NEXT:    pand %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm9, %xmm9
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm9, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    pandn %xmm9, %xmm12
+; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm12
 ; SSE-NEXT:    pxor %xmm9, %xmm9
 ; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
@@ -568,15 +567,14 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm0, %xmm6
 ; SSE-NEXT:    por %xmm5, %xmm6
 ; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm6
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm6
+; SSE-NEXT:    movdqa %xmm10, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm9[8],xmm11[9],xmm9[9],xmm11[10],xmm9[10],xmm11[11],xmm9[11],xmm11[12],xmm9[12],xmm11[13],xmm9[13],xmm11[14],xmm9[14],xmm11[15],xmm9[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,65535,65535,0,65535]
@@ -584,14 +582,14 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm0, %xmm5
 ; SSE-NEXT:    por %xmm1, %xmm5
 ; SSE-NEXT:    packuswb %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm12
-; SSE-NEXT:    pandn %xmm0, %xmm12
-; SSE-NEXT:    pand %xmm10, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm12
+; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm11, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm10
 ; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm15
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
@@ -599,11 +597,11 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    por %xmm3, %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm14[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm3, %xmm8
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
@@ -613,117 +611,118 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm1, %xmm5
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm6, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    pandn %xmm0, %xmm8
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm8
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm9[8],xmm15[9],xmm9[9],xmm15[10],xmm9[10],xmm15[11],xmm9[11],xmm15[12],xmm9[12],xmm15[13],xmm9[13],xmm15[14],xmm9[14],xmm15[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
-; SSE-NEXT:    movaps %xmm0, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm2[0,2]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm15[0,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm5[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    packuswb %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm14, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm9[8],xmm14[9],xmm9[9],xmm14[10],xmm9[10],xmm14[11],xmm9[11],xmm14[12],xmm9[12],xmm14[13],xmm9[13],xmm14[14],xmm9[14],xmm14[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [0,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm6, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pandn %xmm6, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pandn %xmm5, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1],xmm13[2],xmm9[2],xmm13[3],xmm9[3],xmm13[4],xmm9[4],xmm13[5],xmm9[5],xmm13[6],xmm9[6],xmm13[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm5, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    packuswb %xmm5, %xmm0
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    por %xmm3, %xmm10
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,7,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm11, %xmm3
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    pandn %xmm0, %xmm11
+; SSE-NEXT:    por %xmm3, %xmm11
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    packuswb %xmm1, %xmm0
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,1,2,0]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm5
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pand %xmm3, %xmm5
-; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm13, %xmm11
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm13, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    pand %xmm0, %xmm5
-; SSE-NEXT:    por %xmm11, %xmm13
-; SSE-NEXT:    pand %xmm13, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,2,1,3]
+; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3],xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm8, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,0],xmm2[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm2[2,3]
-; SSE-NEXT:    psrlq $48, %xmm2
-; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm2[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm5, %xmm7
+; SSE-NEXT:    pandn %xmm1, %xmm6
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm1[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm1[2,3]
+; SSE-NEXT:    psrlq $48, %xmm1
+; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,2,3]
@@ -738,29 +737,28 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm4, %xmm2
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
 ; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm9[8],xmm13[9],xmm9[9],xmm13[10],xmm9[10],xmm13[11],xmm9[11],xmm13[12],xmm9[12],xmm13[13],xmm9[13],xmm13[14],xmm9[14],xmm13[15],xmm9[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,65535,0,0]
-; SSE-NEXT:    pand %xmm4, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm13[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    pandn %xmm5, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,0,65535,0,0]
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm13[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    pandn %xmm4, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    packuswb %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    packuswb %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rsi)
-; SSE-NEXT:    movdqa %xmm12, (%rdx)
-; SSE-NEXT:    movdqa %xmm1, (%rcx)
-; SSE-NEXT:    movdqa %xmm10, (%r8)
+; SSE-NEXT:    movdqa %xmm12, (%rsi)
+; SSE-NEXT:    movdqa %xmm10, (%rdx)
+; SSE-NEXT:    movdqa %xmm8, (%rcx)
+; SSE-NEXT:    movdqa %xmm11, (%r8)
 ; SSE-NEXT:    movdqa %xmm6, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movdqa %xmm0, (%rax)
@@ -1076,27 +1074,27 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i8_stride6_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $280, %rsp # imm = 0x118
-; SSE-NEXT:    movdqa 64(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    subq $264, %rsp # imm = 0x108
+; SSE-NEXT:    movdqa 64(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm9
-; SSE-NEXT:    movdqa (%rdi), %xmm14
-; SSE-NEXT:    movdqa 16(%rdi), %xmm2
+; SSE-NEXT:    movdqa (%rdi), %xmm12
+; SSE-NEXT:    movdqa 16(%rdi), %xmm14
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pandn %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm5, %xmm15
-; SSE-NEXT:    pand %xmm7, %xmm15
+; SSE-NEXT:    pand %xmm10, %xmm15
 ; SSE-NEXT:    por %xmm0, %xmm15
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
@@ -1107,238 +1105,237 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm11
-; SSE-NEXT:    pand %xmm12, %xmm11
-; SSE-NEXT:    por %xmm1, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm12, %xmm8
+; SSE-NEXT:    pand %xmm11, %xmm8
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,2,1,3]
 ; SSE-NEXT:    pand %xmm3, %xmm1
 ; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm5
 ; SSE-NEXT:    pandn %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm10
-; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    por %xmm0, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,1,2,0]
 ; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rdi), %xmm13
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm13, %xmm0
-; SSE-NEXT:    movdqa 144(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa 160(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm14
-; SSE-NEXT:    pandn %xmm9, %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm9, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    pand %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdi), %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm12
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    pand %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    movdqa 128(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    movdqa 144(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa 160(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    pandn %xmm14, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm12
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    pandn %xmm9, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm11
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm13
+; SSE-NEXT:    movdqa %xmm13, %xmm4
 ; SSE-NEXT:    pand %xmm1, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4
-; SSE-NEXT:    por %xmm8, %xmm4
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 176(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm10
+; SSE-NEXT:    pand %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    pand %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm3, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm14
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pxor %xmm9, %xmm9
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm9[0],xmm15[1],xmm9[1],xmm15[2],xmm9[2],xmm15[3],xmm9[3],xmm15[4],xmm9[4],xmm15[5],xmm9[5],xmm15[6],xmm9[6],xmm15[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm15[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    psrld $16, %xmm0
+; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    pandn %xmm4, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[3,1,2,0]
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    packuswb %xmm4, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1],xmm15[2],xmm4[2],xmm15[3],xmm4[3],xmm15[4],xmm4[4],xmm15[5],xmm4[5],xmm15[6],xmm4[6],xmm15[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm15[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm3[0],xmm14[1],xmm3[1],xmm14[2],xmm3[2],xmm14[3],xmm3[3]
+; SSE-NEXT:    psrld $16, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
-; SSE-NEXT:    packuswb %xmm15, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm5, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm9[8],xmm5[9],xmm9[9],xmm5[10],xmm9[10],xmm5[11],xmm9[11],xmm5[12],xmm9[12],xmm5[13],xmm9[13],xmm5[14],xmm9[14],xmm5[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm3[2],xmm15[3],xmm3[3]
+; SSE-NEXT:    packuswb %xmm15, %xmm14
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    pandn %xmm14, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm14
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm4[8],xmm14[9],xmm4[9],xmm14[10],xmm4[10],xmm14[11],xmm4[11],xmm14[12],xmm4[12],xmm14[13],xmm4[13],xmm14[14],xmm4[14],xmm14[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,65535,0,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm9[0],xmm11[1],xmm9[1],xmm11[2],xmm9[2],xmm11[3],xmm9[3],xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    pand %xmm15, %xmm5
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pand %xmm4, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT:    pandn %xmm14, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3],xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm8[1,3,2,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm14
+; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    packuswb %xmm14, %xmm14
+; SSE-NEXT:    pand %xmm6, %xmm14
+; SSE-NEXT:    por %xmm3, %xmm14
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm4[8],xmm7[9],xmm4[9],xmm7[10],xmm4[10],xmm7[11],xmm4[11],xmm7[12],xmm4[12],xmm7[13],xmm4[13],xmm7[14],xmm4[14],xmm7[15],xmm4[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm3, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3],xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    packuswb %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE-NEXT:    psrld $16, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    pand %xmm15, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm15
-; SSE-NEXT:    por %xmm2, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm15
 ; SSE-NEXT:    packuswb %xmm15, %xmm15
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm15
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT:    pand %xmm6, %xmm15
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm15
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm8
 ; SSE-NEXT:    por %xmm1, %xmm8
 ; SSE-NEXT:    packuswb %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm3, %xmm15
-; SSE-NEXT:    movdqa %xmm3, %xmm8
+; SSE-NEXT:    pand %xmm13, %xmm15
+; SSE-NEXT:    movdqa %xmm13, %xmm7
 ; SSE-NEXT:    por %xmm15, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm4, %xmm12
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pand %xmm5, %xmm10
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
@@ -1346,11 +1343,10 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,5,5,5,5]
@@ -1360,27 +1356,26 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm1, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa (%rsp), %xmm15 # 16-byte Reload
-; SSE-NEXT:    por %xmm1, %xmm15
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pandn %xmm14, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm11[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm4, %xmm13
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm9
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
@@ -1388,12 +1383,11 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm13[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm15, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
@@ -1402,294 +1396,289 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm7, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    pand %xmm7, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm9[8],xmm12[9],xmm9[9],xmm12[10],xmm9[10],xmm12[11],xmm9[11],xmm12[12],xmm9[12],xmm12[13],xmm9[13],xmm12[14],xmm9[14],xmm12[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[3,0]
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm3[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm12[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm12[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm4[8],xmm10[9],xmm4[9],xmm10[10],xmm4[10],xmm10[11],xmm4[11],xmm10[12],xmm4[12],xmm10[13],xmm4[13],xmm10[14],xmm4[14],xmm10[15],xmm4[15]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
+; SSE-NEXT:    movaps %xmm1, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm10[2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm1[3,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm9[8],xmm14[9],xmm9[9],xmm14[10],xmm9[10],xmm14[11],xmm9[11],xmm14[12],xmm9[12],xmm14[13],xmm9[13],xmm14[14],xmm9[14],xmm14[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm1, %xmm5
-; SSE-NEXT:    por %xmm6, %xmm5
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm3[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    pandn %xmm5, %xmm7
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm9[0],xmm15[1],xmm9[1],xmm15[2],xmm9[2],xmm15[3],xmm9[3],xmm15[4],xmm9[4],xmm15[5],xmm9[5],xmm15[6],xmm9[6],xmm15[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm15[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    pand %xmm3, %xmm5
-; SSE-NEXT:    por %xmm7, %xmm5
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    movdqa %xmm8, %xmm15
-; SSE-NEXT:    pandn %xmm5, %xmm15
-; SSE-NEXT:    pand %xmm8, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm15
-; SSE-NEXT:    movdqa %xmm13, %xmm5
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm2[8],xmm13[9],xmm2[9],xmm13[10],xmm2[10],xmm13[11],xmm2[11],xmm13[12],xmm2[12],xmm13[13],xmm2[13],xmm13[14],xmm2[14],xmm13[15],xmm2[15]
-; SSE-NEXT:    movdqa %xmm13, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm5[3,0]
-; SSE-NEXT:    movaps %xmm5, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm6[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm13[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm13[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm1, %xmm7
-; SSE-NEXT:    pandn %xmm5, %xmm1
-; SSE-NEXT:    por %xmm7, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm4[8],xmm12[9],xmm4[9],xmm12[10],xmm4[10],xmm12[11],xmm4[11],xmm12[12],xmm4[12],xmm12[13],xmm4[13],xmm12[14],xmm4[14],xmm12[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,7,5,6,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1],xmm11[2],xmm4[2],xmm11[3],xmm4[3],xmm11[4],xmm4[4],xmm11[5],xmm4[5],xmm11[6],xmm4[6],xmm11[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
+; SSE-NEXT:    movdqa %xmm9, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm3[3,0]
+; SSE-NEXT:    movaps %xmm3, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm9[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm9[2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm6[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,1,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm4[8],xmm13[9],xmm4[9],xmm13[10],xmm4[10],xmm13[11],xmm4[11],xmm13[12],xmm4[12],xmm13[13],xmm4[13],xmm13[14],xmm4[14],xmm13[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm13[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    por %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm5
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
+; SSE-NEXT:    por %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1],xmm10[2],xmm2[2],xmm10[3],xmm2[3],xmm10[4],xmm2[4],xmm10[5],xmm2[5],xmm10[6],xmm2[6],xmm10[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm10[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    pand %xmm3, %xmm5
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    por %xmm5, %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    packuswb %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3],xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm7, %xmm13
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    pandn %xmm1, %xmm13
+; SSE-NEXT:    por %xmm0, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,1,2,0]
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm3
+; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm13, %xmm12
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pand %xmm12, %xmm11
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,2,1,3]
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm6
+; SSE-NEXT:    packuswb %xmm0, %xmm5
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,0]
 ; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
+; SSE-NEXT:    pandn %xmm5, %xmm8
 ; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm8
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm5, %xmm6
-; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    packuswb %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,1,2,0]
+; SSE-NEXT:    pand %xmm15, %xmm3
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm3[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm7, %xmm7
-; SSE-NEXT:    pand %xmm5, %xmm7
-; SSE-NEXT:    por %xmm6, %xmm7
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm3[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    pand %xmm2, %xmm6
+; SSE-NEXT:    por %xmm5, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm13, %xmm3
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm3, %xmm13
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pand %xmm13, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
+; SSE-NEXT:    pand %xmm12, %xmm3
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm3, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,2,1,3]
+; SSE-NEXT:    pand %xmm15, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm6
+; SSE-NEXT:    packuswb %xmm3, %xmm5
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3],xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm9, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm6[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm6[2,3]
-; SSE-NEXT:    psrlq $48, %xmm6
-; SSE-NEXT:    psrldq {{.*#+}} xmm7 = xmm7[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm9[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm7, %xmm6
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    pandn %xmm6, %xmm7
-; SSE-NEXT:    movdqa %xmm10, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm6[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm11
-; SSE-NEXT:    pandn %xmm9, %xmm11
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1],xmm10[2],xmm2[2],xmm10[3],xmm2[3],xmm10[4],xmm2[4],xmm10[5],xmm2[5],xmm10[6],xmm2[6],xmm10[7],xmm2[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm10[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm10
-; SSE-NEXT:    por %xmm11, %xmm10
-; SSE-NEXT:    packuswb %xmm10, %xmm10
-; SSE-NEXT:    pand %xmm5, %xmm10
-; SSE-NEXT:    por %xmm7, %xmm10
-; SSE-NEXT:    movdqa %xmm12, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm2[8],xmm12[9],xmm2[9],xmm12[10],xmm2[10],xmm12[11],xmm2[11],xmm12[12],xmm2[12],xmm12[13],xmm2[13],xmm12[14],xmm2[14],xmm12[15],xmm2[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm12[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm9[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,0,65535,0,0]
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    pandn %xmm11, %xmm12
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3],xmm7[4],xmm2[4],xmm7[5],xmm2[5],xmm7[6],xmm2[6],xmm7[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm9, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm12
-; SSE-NEXT:    packuswb %xmm12, %xmm11
-; SSE-NEXT:    movdqa %xmm0, %xmm7
-; SSE-NEXT:    pandn %xmm11, %xmm7
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm7
-; SSE-NEXT:    movdqa %xmm14, %xmm10
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm2[8],xmm10[9],xmm2[9],xmm10[10],xmm2[10],xmm10[11],xmm2[11],xmm10[12],xmm2[12],xmm10[13],xmm2[13],xmm10[14],xmm2[14],xmm10[15],xmm2[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1],xmm14[2],xmm2[2],xmm14[3],xmm2[3],xmm14[4],xmm2[4],xmm14[5],xmm2[5],xmm14[6],xmm2[6],xmm14[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm14, %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm10[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0],xmm10[2,3]
-; SSE-NEXT:    psrlq $48, %xmm10
-; SSE-NEXT:    psrldq {{.*#+}} xmm11 = xmm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm14[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    pandn %xmm5, %xmm3
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm5[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[2,3]
+; SSE-NEXT:    psrlq $48, %xmm5
+; SSE-NEXT:    psrldq {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm9, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm5[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm10
+; SSE-NEXT:    pandn %xmm7, %xmm10
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3],xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm9[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm7[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm5, %xmm9
+; SSE-NEXT:    por %xmm10, %xmm9
+; SSE-NEXT:    packuswb %xmm9, %xmm9
+; SSE-NEXT:    pand %xmm2, %xmm9
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    movdqa %xmm11, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm4[8],xmm11[9],xmm4[9],xmm11[10],xmm4[10],xmm11[11],xmm4[11],xmm11[12],xmm4[12],xmm11[13],xmm4[13],xmm11[14],xmm4[14],xmm11[15],xmm4[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm11[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm7[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,0,65535,0,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm11
+; SSE-NEXT:    pandn %xmm10, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3],xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm7, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm11
 ; SSE-NEXT:    packuswb %xmm11, %xmm10
-; SSE-NEXT:    movdqa %xmm1, %xmm11
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm2[8],xmm11[9],xmm2[9],xmm11[10],xmm2[10],xmm11[11],xmm2[11],xmm11[12],xmm2[12],xmm11[13],xmm2[13],xmm11[14],xmm2[14],xmm11[15],xmm2[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm12
-; SSE-NEXT:    pandn %xmm11, %xmm6
-; SSE-NEXT:    por %xmm12, %xmm6
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm10, %xmm6
+; SSE-NEXT:    pand %xmm0, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm6
+; SSE-NEXT:    movdqa %xmm14, %xmm11
+; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1],xmm11[2],xmm4[2],xmm11[3],xmm4[3],xmm11[4],xmm4[4],xmm11[5],xmm4[5],xmm11[6],xmm4[6],xmm11[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm9[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm9[2,3]
+; SSE-NEXT:    psrlq $48, %xmm9
+; SSE-NEXT:    psrldq {{.*#+}} xmm10 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm10, %xmm9
+; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm4[8],xmm10[9],xmm4[9],xmm10[10],xmm4[10],xmm10[11],xmm4[11],xmm10[12],xmm4[12],xmm10[13],xmm4[13],xmm10[14],xmm4[14],xmm10[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm1[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm5, %xmm11
 ; SSE-NEXT:    pandn %xmm10, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm13, %xmm5
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm2[8],xmm13[9],xmm2[9],xmm13[10],xmm2[10],xmm13[11],xmm2[11],xmm13[12],xmm2[12],xmm13[13],xmm2[13],xmm13[14],xmm2[14],xmm13[15],xmm2[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm9, %xmm4
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm13[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    pandn %xmm5, %xmm9
-; SSE-NEXT:    por %xmm4, %xmm9
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    packuswb %xmm9, %xmm4
-; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, (%rsi)
+; SSE-NEXT:    por %xmm11, %xmm5
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm4[8],xmm12[9],xmm4[9],xmm12[10],xmm4[10],xmm12[11],xmm4[11],xmm12[12],xmm4[12],xmm12[13],xmm4[13],xmm12[14],xmm4[14],xmm12[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm12[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    pandn %xmm4, %xmm7
+; SSE-NEXT:    por %xmm2, %xmm7
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    packuswb %xmm7, %xmm2
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, (%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, 16(%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -1698,15 +1687,15 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movaps %xmm1, 16(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, (%rcx)
+; SSE-NEXT:    movdqa %xmm13, 16(%r8)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%r8)
-; SSE-NEXT:    movdqa %xmm15, (%r8)
+; SSE-NEXT:    movaps %xmm1, (%r8)
 ; SSE-NEXT:    movdqa %xmm3, 16(%r9)
 ; SSE-NEXT:    movdqa %xmm8, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movdqa %xmm0, 16(%rax)
-; SSE-NEXT:    movdqa %xmm7, (%rax)
-; SSE-NEXT:    addq $280, %rsp # imm = 0x118
+; SSE-NEXT:    movdqa %xmm6, (%rax)
+; SSE-NEXT:    addq $264, %rsp # imm = 0x108
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: load_i8_stride6_vf32:
@@ -2299,7 +2288,7 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i8_stride6_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $808, %rsp # imm = 0x328
+; SSE-NEXT:    subq $792, %rsp # imm = 0x318
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm5
@@ -2311,22 +2300,22 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,0,65535,65535,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm11
+; SSE-NEXT:    movdqa %xmm1, %xmm10
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
@@ -2339,9 +2328,8 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm3, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
+; SSE-NEXT:    pand %xmm10, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
@@ -2351,15 +2339,13 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    pand %xmm9, %xmm4
+; SSE-NEXT:    pand %xmm13, %xmm4
 ; SSE-NEXT:    por %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,1,2,0]
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
@@ -2372,282 +2358,272 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 320(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 336(%rdi), %xmm13
+; SSE-NEXT:    movdqa 336(%rdi), %xmm12
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    pandn %xmm12, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    pandn %xmm12, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm13
-; SSE-NEXT:    por %xmm0, %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm12
+; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa 304(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    movdqa 304(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm7
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 288(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    movdqa 288(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,2,1,3]
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa 368(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa 352(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
+; SSE-NEXT:    movdqa 352(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,0]
+; SSE-NEXT:    pand %xmm10, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    movdqa %xmm4, %xmm9
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 224(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 240(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm6, %xmm14
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    movdqa 240(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm13, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm5
 ; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa 208(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    movdqa 208(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,1,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm10
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa 272(%rdi), %xmm15
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pandn %xmm15, %xmm2
-; SSE-NEXT:    movdqa 256(%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm6, %xmm12
-; SSE-NEXT:    por %xmm2, %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
+; SSE-NEXT:    movdqa 272(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    pandn %xmm14, %xmm2
+; SSE-NEXT:    movdqa 256(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm13, %xmm15
+; SSE-NEXT:    por %xmm2, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[3,1,2,0]
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa 144(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pandn %xmm10, %xmm4
+; SSE-NEXT:    movdqa 144(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    pandn %xmm9, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    pandn %xmm9, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm6, %xmm10
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm9
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pandn %xmm7, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm13, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    movdqa 112(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm8, %xmm7
-; SSE-NEXT:    movdqa %xmm8, %xmm9
-; SSE-NEXT:    pandn %xmm6, %xmm9
-; SSE-NEXT:    movdqa 160(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    pandn %xmm6, %xmm8
+; SSE-NEXT:    movdqa 160(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm7
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm13, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    pandn %xmm14, %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    pandn %xmm15, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm15
-; SSE-NEXT:    movdqa 96(%rdi), %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 176(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm13
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm5, %xmm1
+; SSE-NEXT:    pandn %xmm4, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    por %xmm9, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm7, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
+; SSE-NEXT:    por %xmm8, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm1, %xmm5
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    por %xmm11, %xmm5
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    pand %xmm8, %xmm5
+; SSE-NEXT:    por %xmm10, %xmm5
+; SSE-NEXT:    pandn %xmm14, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,1,2,0]
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm5
 ; SSE-NEXT:    por %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pxor %xmm5, %xmm5
@@ -2658,11 +2634,11 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,2,3,3]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
 ; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
-; SSE-NEXT:    packuswb %xmm15, %xmm4
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; SSE-NEXT:    packuswb %xmm14, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm2, %xmm4
@@ -2671,48 +2647,48 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm15
-; SSE-NEXT:    pandn %xmm4, %xmm15
+; SSE-NEXT:    movdqa %xmm0, %xmm14
+; SSE-NEXT:    pandn %xmm4, %xmm14
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,0,4,5,6,7]
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm15, %xmm4
+; SSE-NEXT:    por %xmm14, %xmm4
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm9, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,7,6,4]
-; SSE-NEXT:    pand %xmm2, %xmm15
-; SSE-NEXT:    por %xmm1, %xmm15
-; SSE-NEXT:    packuswb %xmm15, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm15
-; SSE-NEXT:    pandn %xmm1, %xmm15
-; SSE-NEXT:    pand %xmm7, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm6[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,7,6,4]
+; SSE-NEXT:    pand %xmm2, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm14
+; SSE-NEXT:    packuswb %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm14
+; SSE-NEXT:    pandn %xmm1, %xmm14
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm5[0],xmm13[1],xmm5[1],xmm13[2],xmm5[2],xmm13[3],xmm5[3],xmm13[4],xmm5[4],xmm13[5],xmm5[5],xmm13[6],xmm5[6],xmm13[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,2,3,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3],xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[2,2,3,3]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
 ; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm1[2],xmm13[3],xmm1[3]
-; SSE-NEXT:    packuswb %xmm13, %xmm4
-; SSE-NEXT:    movdqa %xmm9, %xmm15
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm1[2],xmm12[3],xmm1[3]
+; SSE-NEXT:    packuswb %xmm12, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm14
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm6, %xmm4
@@ -2720,46 +2696,46 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    movdqa %xmm0, %xmm13
-; SSE-NEXT:    pandn %xmm4, %xmm13
+; SSE-NEXT:    movdqa %xmm0, %xmm12
+; SSE-NEXT:    pandn %xmm4, %xmm12
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,0,4,5,6,7]
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm13, %xmm4
+; SSE-NEXT:    por %xmm12, %xmm4
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm9, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm2, %xmm13
-; SSE-NEXT:    pandn %xmm1, %xmm13
+; SSE-NEXT:    movdqa %xmm2, %xmm12
+; SSE-NEXT:    pandn %xmm1, %xmm12
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    por %xmm13, %xmm1
+; SSE-NEXT:    por %xmm12, %xmm1
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm13
-; SSE-NEXT:    pandn %xmm1, %xmm13
-; SSE-NEXT:    pand %xmm7, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm13
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm1, %xmm12
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[2,2,3,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3],xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[2,2,3,3]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
 ; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm14[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm1[2],xmm9[3],xmm1[3]
-; SSE-NEXT:    packuswb %xmm9, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm1[2],xmm8[3],xmm1[3]
+; SSE-NEXT:    packuswb %xmm8, %xmm4
+; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm6, %xmm4
@@ -2767,41 +2743,41 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    movdqa %xmm0, %xmm9
-; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    movdqa %xmm0, %xmm8
+; SSE-NEXT:    pandn %xmm4, %xmm8
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,0,4,5,6,7]
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm9, %xmm4
+; SSE-NEXT:    por %xmm8, %xmm4
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm15, %xmm4
+; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm2, %xmm9
-; SSE-NEXT:    pandn %xmm1, %xmm9
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm5[8],xmm12[9],xmm5[9],xmm12[10],xmm5[10],xmm12[11],xmm5[11],xmm12[12],xmm5[12],xmm12[13],xmm5[13],xmm12[14],xmm5[14],xmm12[15],xmm5[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm5[8],xmm15[9],xmm5[9],xmm15[10],xmm5[10],xmm15[11],xmm5[11],xmm15[12],xmm5[12],xmm15[13],xmm5[13],xmm15[14],xmm5[14],xmm15[15],xmm5[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm15[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    por %xmm9, %xmm1
+; SSE-NEXT:    por %xmm8, %xmm1
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm9
-; SSE-NEXT:    pandn %xmm1, %xmm9
-; SSE-NEXT:    pand %xmm7, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3],xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[2,2,3,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[2,2,3,3]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
 ; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm10[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
 ; SSE-NEXT:    packuswb %xmm6, %xmm4
@@ -2818,32 +2794,34 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    pandn %xmm4, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm5[8],xmm7[9],xmm5[9],xmm7[10],xmm5[10],xmm7[11],xmm5[11],xmm7[12],xmm5[12],xmm7[13],xmm5[13],xmm7[14],xmm5[14],xmm7[15],xmm5[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm11
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pand %xmm14, %xmm4
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
@@ -2853,719 +2831,712 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm12
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm10
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm11, %xmm3
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pand %xmm14, %xmm13
+; SSE-NEXT:    movdqa %xmm14, %xmm7
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm8
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pand %xmm7, %xmm10
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    packuswb %xmm2, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm12, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm7
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0]
-; SSE-NEXT:    movaps %xmm0, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm1[8],xmm4[9],xmm1[9],xmm4[10],xmm1[10],xmm4[11],xmm1[11],xmm4[12],xmm1[12],xmm4[13],xmm1[13],xmm4[14],xmm1[14],xmm4[15],xmm1[15]
+; SSE-NEXT:    pxor %xmm7, %xmm7
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[0,2]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm4[0,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm1
+; SSE-NEXT:    packuswb %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3],xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm14, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm5[8],xmm12[9],xmm5[9],xmm12[10],xmm5[10],xmm12[11],xmm5[11],xmm12[12],xmm5[12],xmm12[13],xmm5[13],xmm12[14],xmm5[14],xmm12[15],xmm5[15]
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[3,0]
-; SSE-NEXT:    movaps %xmm2, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm4[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm12[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm12[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm5[8],xmm15[9],xmm5[9],xmm15[10],xmm5[10],xmm15[11],xmm5[11],xmm15[12],xmm5[12],xmm15[13],xmm5[13],xmm15[14],xmm5[14],xmm15[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    pand %xmm3, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    por %xmm4, %xmm0
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm7[8],xmm0[9],xmm7[9],xmm0[10],xmm7[10],xmm0[11],xmm7[11],xmm0[12],xmm7[12],xmm0[13],xmm7[13],xmm0[14],xmm7[14],xmm0[15],xmm7[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
+; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm14, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[3,0]
-; SSE-NEXT:    movaps %xmm2, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm4[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm8[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm8[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm5[8],xmm13[9],xmm5[9],xmm13[10],xmm5[10],xmm13[11],xmm5[11],xmm13[12],xmm5[12],xmm13[13],xmm5[13],xmm13[14],xmm5[14],xmm13[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm6
 ; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm1[8],xmm13[9],xmm1[9],xmm13[10],xmm1[10],xmm13[11],xmm1[11],xmm13[12],xmm1[12],xmm13[13],xmm1[13],xmm13[14],xmm1[14],xmm13[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
+; SSE-NEXT:    movaps %xmm0, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm2[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm13[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm13[2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    pand %xmm3, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3],xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm14, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm5[8],xmm7[9],xmm5[9],xmm7[10],xmm5[10],xmm7[11],xmm5[11],xmm7[12],xmm5[12],xmm7[13],xmm5[13],xmm7[14],xmm5[14],xmm7[15],xmm5[15]
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[3,0]
-; SSE-NEXT:    movaps %xmm2, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm4[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm7[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm7[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm5[8],xmm11[9],xmm5[9],xmm11[10],xmm5[10],xmm11[11],xmm5[11],xmm11[12],xmm5[12],xmm11[13],xmm5[13],xmm11[14],xmm5[14],xmm11[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm11[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm1[8],xmm14[9],xmm1[9],xmm14[10],xmm1[10],xmm14[11],xmm1[11],xmm14[12],xmm1[12],xmm14[13],xmm1[13],xmm14[14],xmm1[14],xmm14[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm4
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3],xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pand %xmm14, %xmm3
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3],xmm11[4],xmm1[4],xmm11[5],xmm1[5],xmm11[6],xmm1[6],xmm11[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
+; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm6, %xmm9
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    pand %xmm8, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm1[8],xmm10[9],xmm1[9],xmm10[10],xmm1[10],xmm10[11],xmm1[11],xmm10[12],xmm1[12],xmm10[13],xmm1[13],xmm10[14],xmm1[14],xmm10[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
+; SSE-NEXT:    movaps %xmm0, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm2[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm1[8],xmm7[9],xmm1[9],xmm7[10],xmm1[10],xmm7[11],xmm1[11],xmm7[12],xmm1[12],xmm7[13],xmm1[13],xmm7[14],xmm1[14],xmm7[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm8, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1],xmm8[2],xmm1[2],xmm8[3],xmm1[3],xmm8[4],xmm1[4],xmm8[5],xmm1[5],xmm8[6],xmm1[6],xmm8[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm2[3,0]
+; SSE-NEXT:    movaps %xmm2, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm6[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,1,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm1[8],xmm15[9],xmm1[9],xmm15[10],xmm1[10],xmm15[11],xmm1[11],xmm15[12],xmm1[12],xmm15[13],xmm1[13],xmm15[14],xmm1[14],xmm15[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm15[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm3, %xmm7
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    por %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm5, %xmm6
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,5]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3],xmm12[4],xmm0[4],xmm12[5],xmm0[5],xmm12[6],xmm0[6],xmm12[7],xmm0[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm5
+; SSE-NEXT:    packuswb %xmm4, %xmm2
+; SSE-NEXT:    pandn %xmm2, %xmm9
+; SSE-NEXT:    por %xmm5, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,1,2,0]
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm1[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pand %xmm11, %xmm13
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,2,1,3]
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm15
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    pand %xmm8, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; SSE-NEXT:    packuswb %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0]
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm2[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm7
+; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    por %xmm6, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm3
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
+; SSE-NEXT:    pand %xmm10, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    packuswb %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm13, %xmm6
-; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
+; SSE-NEXT:    packuswb %xmm7, %xmm6
+; SSE-NEXT:    movdqa %xmm15, %xmm7
+; SSE-NEXT:    pandn %xmm6, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    pandn %xmm2, %xmm7
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm7
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm2[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[2,3]
-; SSE-NEXT:    psrlq $48, %xmm2
-; SSE-NEXT:    psrldq {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm8
-; SSE-NEXT:    pandn %xmm2, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[3,1,2,0]
+; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm8, %xmm8
+; SSE-NEXT:    pand %xmm15, %xmm8
+; SSE-NEXT:    por %xmm7, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm2
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[0,2,1,3]
+; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm6, %xmm7
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,0,2]
+; SSE-NEXT:    packuswb %xmm8, %xmm7
+; SSE-NEXT:    movdqa %xmm15, %xmm8
+; SSE-NEXT:    pandn %xmm7, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[3,1,2,0]
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm7[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm9, %xmm9
+; SSE-NEXT:    pand %xmm15, %xmm9
+; SSE-NEXT:    por %xmm8, %xmm9
 ; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm10
-; SSE-NEXT:    pandn %xmm2, %xmm10
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3],xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm11[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    por %xmm10, %xmm6
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm13, %xmm6
-; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm5[8],xmm9[9],xmm5[9],xmm9[10],xmm5[10],xmm9[11],xmm5[11],xmm9[12],xmm5[12],xmm9[13],xmm5[13],xmm9[14],xmm5[14],xmm9[15],xmm5[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm9[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,65535,0,65535,0,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm7
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm7, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm2[0,2,1,3]
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm8
+; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    pandn %xmm8, %xmm7
+; SSE-NEXT:    pand %xmm1, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm7
+; SSE-NEXT:    movdqa %xmm0, %xmm8
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm8[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm8[2,3]
+; SSE-NEXT:    psrlq $48, %xmm8
+; SSE-NEXT:    psrldq {{.*#+}} xmm9 = xmm9[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm9, %xmm8
+; SSE-NEXT:    movdqa %xmm15, %xmm10
 ; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm10
-; SSE-NEXT:    packuswb %xmm10, %xmm8
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm9
-; SSE-NEXT:    movdqa %xmm12, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm9, %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm6[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm6[2,3]
-; SSE-NEXT:    psrlq $48, %xmm6
-; SSE-NEXT:    psrldq {{.*#+}} xmm8 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm9[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm8, %xmm6
+; SSE-NEXT:    movdqa %xmm12, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm11
+; SSE-NEXT:    pandn %xmm8, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3],xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm12[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm8[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    por %xmm11, %xmm12
+; SSE-NEXT:    packuswb %xmm12, %xmm12
+; SSE-NEXT:    pand %xmm15, %xmm12
+; SSE-NEXT:    por %xmm10, %xmm12
 ; SSE-NEXT:    movdqa %xmm13, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm0, %xmm10
-; SSE-NEXT:    pandn %xmm6, %xmm10
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    por %xmm10, %xmm6
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm13, %xmm6
-; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm1[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm5[8],xmm13[9],xmm5[9],xmm13[10],xmm5[10],xmm13[11],xmm5[11],xmm13[12],xmm5[12],xmm13[13],xmm5[13],xmm13[14],xmm5[14],xmm13[15],xmm5[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm13[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa %xmm11, %xmm12
-; SSE-NEXT:    pandn %xmm10, %xmm12
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,65535,0,65535,0,0]
+; SSE-NEXT:    movdqa %xmm11, %xmm13
+; SSE-NEXT:    pandn %xmm10, %xmm13
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3],xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,1,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    pand %xmm11, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm12
-; SSE-NEXT:    packuswb %xmm12, %xmm8
-; SSE-NEXT:    movdqa %xmm4, %xmm12
-; SSE-NEXT:    pandn %xmm8, %xmm12
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm9, %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm6[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm6[2,3]
-; SSE-NEXT:    psrlq $48, %xmm6
-; SSE-NEXT:    psrldq {{.*#+}} xmm8 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm9[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm8, %xmm6
-; SSE-NEXT:    movdqa %xmm13, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    movdqa %xmm14, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm0, %xmm9
-; SSE-NEXT:    pandn %xmm6, %xmm9
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm14[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm6[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    por %xmm9, %xmm10
-; SSE-NEXT:    packuswb %xmm10, %xmm10
-; SSE-NEXT:    pand %xmm13, %xmm10
-; SSE-NEXT:    por %xmm8, %xmm10
-; SSE-NEXT:    movdqa %xmm15, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm5[8],xmm15[9],xmm5[9],xmm15[10],xmm5[10],xmm15[11],xmm5[11],xmm15[12],xmm5[12],xmm15[13],xmm5[13],xmm15[14],xmm5[14],xmm15[15],xmm5[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm15[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa %xmm11, %xmm9
-; SSE-NEXT:    pandn %xmm8, %xmm9
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm9
-; SSE-NEXT:    packuswb %xmm9, %xmm8
-; SSE-NEXT:    movdqa %xmm4, %xmm6
-; SSE-NEXT:    pandn %xmm8, %xmm6
-; SSE-NEXT:    pand %xmm4, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm6
-; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    por %xmm8, %xmm13
+; SSE-NEXT:    packuswb %xmm13, %xmm10
 ; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm1, %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm8[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm8[2,3]
-; SSE-NEXT:    psrlq $48, %xmm8
-; SSE-NEXT:    psrldq {{.*#+}} xmm9 = xmm9[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm9, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm10, %xmm8
+; SSE-NEXT:    pand %xmm1, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm8
 ; SSE-NEXT:    movdqa %xmm14, %xmm9
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm5[8],xmm9[9],xmm5[9],xmm9[10],xmm5[10],xmm9[11],xmm5[11],xmm9[12],xmm5[12],xmm9[13],xmm5[13],xmm9[14],xmm5[14],xmm9[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm14[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa %xmm14, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
+; SSE-NEXT:    movdqa %xmm9, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm10[2,3]
+; SSE-NEXT:    psrlq $48, %xmm10
+; SSE-NEXT:    psrldq {{.*#+}} xmm12 = xmm12[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm12, %xmm10
+; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    pandn %xmm10, %xmm12
+; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm0, %xmm14
+; SSE-NEXT:    pandn %xmm10, %xmm14
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm4[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    por %xmm10, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm10[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm13
+; SSE-NEXT:    por %xmm14, %xmm13
+; SSE-NEXT:    packuswb %xmm13, %xmm13
+; SSE-NEXT:    pand %xmm15, %xmm13
+; SSE-NEXT:    por %xmm12, %xmm13
+; SSE-NEXT:    movdqa %xmm3, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    movdqa %xmm11, %xmm14
+; SSE-NEXT:    pandn %xmm12, %xmm14
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3],xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm11, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm14
+; SSE-NEXT:    packuswb %xmm14, %xmm10
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm10, %xmm12
+; SSE-NEXT:    pand %xmm1, %xmm13
+; SSE-NEXT:    por %xmm13, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm10
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm3[8],xmm10[9],xmm3[9],xmm10[10],xmm3[10],xmm10[11],xmm3[11],xmm10[12],xmm3[12],xmm10[13],xmm3[13],xmm10[14],xmm3[14],xmm10[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1],xmm9[2],xmm3[2],xmm9[3],xmm3[3],xmm9[4],xmm3[4],xmm9[5],xmm3[5],xmm9[6],xmm3[6],xmm9[7],xmm3[7]
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    movdqa %xmm9, %xmm13
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm10[2,3]
+; SSE-NEXT:    psrlq $48, %xmm10
+; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm13, %xmm10
+; SSE-NEXT:    movdqa %xmm15, %xmm13
+; SSE-NEXT:    pandn %xmm10, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm4[8],xmm10[9],xmm4[9],xmm10[10],xmm4[10],xmm10[11],xmm4[11],xmm10[12],xmm4[12],xmm10[13],xmm4[13],xmm10[14],xmm4[14],xmm10[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    pandn %xmm10, %xmm9
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm10[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm14
+; SSE-NEXT:    por %xmm9, %xmm14
+; SSE-NEXT:    packuswb %xmm14, %xmm14
+; SSE-NEXT:    pand %xmm15, %xmm14
+; SSE-NEXT:    por %xmm13, %xmm14
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    movdqa %xmm11, %xmm13
+; SSE-NEXT:    pandn %xmm10, %xmm13
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3],xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm13
+; SSE-NEXT:    packuswb %xmm13, %xmm9
+; SSE-NEXT:    movdqa %xmm1, %xmm13
+; SSE-NEXT:    pandn %xmm9, %xmm13
+; SSE-NEXT:    pand %xmm1, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm13
+; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm3, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm9[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm9[2,3]
+; SSE-NEXT:    psrlq $48, %xmm9
+; SSE-NEXT:    psrldq {{.*#+}} xmm10 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm10, %xmm9
+; SSE-NEXT:    movdqa %xmm6, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm6[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm14
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    por %xmm14, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm13
-; SSE-NEXT:    por %xmm13, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm5[8],xmm14[9],xmm5[9],xmm14[10],xmm5[10],xmm14[11],xmm5[11],xmm14[12],xmm5[12],xmm14[13],xmm5[13],xmm14[14],xmm5[14],xmm14[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm14[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm15
+; SSE-NEXT:    por %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm2[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,7,4]
 ; SSE-NEXT:    pandn %xmm5, %xmm11
-; SSE-NEXT:    por %xmm1, %xmm11
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    packuswb %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm11
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    packuswb %xmm11, %xmm4
+; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -3606,11 +3577,11 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm4, 16(%rax)
-; SSE-NEXT:    movdqa %xmm6, 32(%rax)
+; SSE-NEXT:    movdqa %xmm1, 16(%rax)
+; SSE-NEXT:    movdqa %xmm13, 32(%rax)
 ; SSE-NEXT:    movdqa %xmm12, 48(%rax)
-; SSE-NEXT:    movdqa %xmm2, (%rax)
-; SSE-NEXT:    addq $808, %rsp # imm = 0x328
+; SSE-NEXT:    movdqa %xmm8, (%rax)
+; SSE-NEXT:    addq $792, %rsp # imm = 0x318
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: load_i8_stride6_vf64:

diff  --git a/llvm/test/CodeGen/X86/vector-shift-ashr-128.ll b/llvm/test/CodeGen/X86/vector-shift-ashr-128.ll
index 938fba0490b557..e6b5998d965f06 100644
--- a/llvm/test/CodeGen/X86/vector-shift-ashr-128.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-ashr-128.ll
@@ -1160,9 +1160,9 @@ define <16 x i8> @splatvar_modulo_shift_v16i8(<16 x i8> %a, <16 x i8> %b) nounwi
 ;
 ; XOPAVX1-LABEL: splatvar_modulo_shift_v16i8:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
 ; XOPAVX1-NEXT:    vpshab %xmm1, %xmm0, %xmm0
 ; XOPAVX1-NEXT:    retq

diff  --git a/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll b/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll
index 5a70e5d4a2b560..6d19a81d9fd782 100644
--- a/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll
@@ -1265,9 +1265,9 @@ define <32 x i8> @splatvar_modulo_shift_v32i8(<32 x i8> %a, <32 x i8> %b) nounwi
 ;
 ; XOPAVX1-LABEL: splatvar_modulo_shift_v32i8:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; XOPAVX1-NEXT:    vpshab %xmm1, %xmm2, %xmm2

diff  --git a/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
index e248aafab52557..7a2dcd1c8ca8c9 100644
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
@@ -941,9 +941,9 @@ define <16 x i8> @splatvar_modulo_shift_v16i8(<16 x i8> %a, <16 x i8> %b) nounwi
 ;
 ; XOPAVX1-LABEL: splatvar_modulo_shift_v16i8:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
 ; XOPAVX1-NEXT:    vpshlb %xmm1, %xmm0, %xmm0
 ; XOPAVX1-NEXT:    retq

diff  --git a/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
index 77f5f2660af7ed..5fe661c7e77780 100644
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
@@ -1037,9 +1037,9 @@ define <32 x i8> @splatvar_modulo_shift_v32i8(<32 x i8> %a, <32 x i8> %b) nounwi
 ;
 ; XOPAVX1-LABEL: splatvar_modulo_shift_v32i8:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; XOPAVX1-NEXT:    vpshlb %xmm1, %xmm2, %xmm2

diff  --git a/llvm/test/CodeGen/X86/vector-shift-shl-128.ll b/llvm/test/CodeGen/X86/vector-shift-shl-128.ll
index deb1514e42c4a2..12f971fb83b56f 100644
--- a/llvm/test/CodeGen/X86/vector-shift-shl-128.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-shl-128.ll
@@ -848,9 +848,9 @@ define <16 x i8> @splatvar_modulo_shift_v16i8(<16 x i8> %a, <16 x i8> %b) nounwi
 ;
 ; XOPAVX1-LABEL: splatvar_modulo_shift_v16i8:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpshlb %xmm1, %xmm0, %xmm0
 ; XOPAVX1-NEXT:    retq
 ;

diff  --git a/llvm/test/CodeGen/X86/vector-shift-shl-256.ll b/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
index c355eeaa42b669..76944994c87d17 100644
--- a/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
@@ -962,9 +962,9 @@ define <32 x i8> @splatvar_modulo_shift_v32i8(<32 x i8> %a, <32 x i8> %b) nounwi
 ;
 ; XOPAVX1-LABEL: splatvar_modulo_shift_v32i8:
 ; XOPAVX1:       # %bb.0:
-; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; XOPAVX1-NEXT:    vpshlb %xmm1, %xmm2, %xmm2
 ; XOPAVX1-NEXT:    vpshlb %xmm1, %xmm0, %xmm0


        


More information about the llvm-commits mailing list