[x86] split more v8f32/v8i32 shuffles in lowering

[spatel]

Summary

Similar to D57867 - this is a 1-line patch with lots of test diffs.
In most cases with half-vector-width narrowing potential, using an extract + 128-bit vshufps is a win because it replaces a 256-bit shuffle with a 128-bit shufle.

There's 1 potentially controversial diff pattern for a target with "fast-variable-shuffle".
We are changing:

vmovaps {{.*#+}} ymm1 = [load 256-bit constant permute mask]
vpermps %ymm0, %ymm1, %ymm0

to:

vextractf128 $1, %ymm0, %xmm1
vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]

That could be a regression if the permute mask load could be moved out of a loop and the 256-bit op is executed at the same speed/power as a 128-bit op. But I think the extract+shufps combo is the right default choice at this level because it removes a ymm instruction. We should form 256-bit vpermps from the extract+shufps as a later optimization within a loop if that would be profitable.

[spatel]

Better to stage this in 2 parts?
I can add a check for fast-variable-shuffle, so we get the clear improvements. Then, a follow-up can remove that check and see if that results in any real-world fallout.

[RKSimon]

In D58181#1399533, @spatel wrote:

Better to stage this in 2 parts?
I can add a check for fast-variable-shuffle, so we get the clear improvements. Then, a follow-up can remove that check and see if that results in any real-world fallout.

Yes, please can you update this patch for just the slow path?

[spatel]

Patch updated:
Restrict the change to targets without fast-variable-shuffle.

[RKSimon]

LGTM - cheers.

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

if (!UndefLower) {
     // Always extract lowers when setting lower - these are all free subreg ops.
     if (NumUpperHalves == 0)
       return getShuffleHalfVectors(DL, V1, V2, HalfMask, HalfIdx1, HalfIdx2,
                                    UndefLower, DAG);
 
     if (NumUpperHalves == 1) {
       // AVX2 has efficient 32/64-bit element cross-lane shuffles.
       if (Subtarget.hasAVX2()) {
-        // extract128 + vunpckhps, is better than vblend + vpermps.
-        // TODO: Refine to account for unary shuffle, splat, and other masks?
-        if (EltWidth == 32 && NumLowerHalves &&
-            HalfVT.is128BitVector() && !is128BitUnpackShuffleMask(HalfMask))
+        // extract128 + vunpckhps/vshufps, is better than vblend + vpermps.
+        if (EltWidth == 32 && NumLowerHalves && HalfVT.is128BitVector() &&
+            !is128BitUnpackShuffleMask(HalfMask) &&
+            (!isSingleSHUFPSMask(HalfMask) ||
+             Subtarget.hasFastVariableShuffle()))
           return SDValue();
         // If this is a unary shuffle (assume that the 2nd operand is
         // canonicalized to undef), then we can use vpermpd. Otherwise, we
         // are better off extracting the upper half of 1 operand and using a
         // narrow shuffle.
         if (EltWidth == 64 && V2.isUndef())
           return SDValue();
       }

llvm/trunk/test/CodeGen/X86/avx2-conversions.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=X32,X32-SLOW
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefixes=X32,X32-FAST
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=X64,X64-SLOW
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefixes=X64,X64-FAST
 
 define <4 x i32> @trunc4(<4 x i64> %A) nounwind {
 ; X32-SLOW-LABEL: trunc4:
 ; X32-SLOW:       # %bb.0:
-; X32-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; X32-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; X32-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; X32-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; X32-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; X32-SLOW-NEXT:    vzeroupper
 ; X32-SLOW-NEXT:    retl
 ;
 ; X32-FAST-LABEL: trunc4:
 ; X32-FAST:       # %bb.0:
 ; X32-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; X32-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
 ; X32-FAST-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; X32-FAST-NEXT:    vzeroupper
 ; X32-FAST-NEXT:    retl
 ;
 ; X64-SLOW-LABEL: trunc4:
 ; X64-SLOW:       # %bb.0:
-; X64-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; X64-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; X64-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; X64-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; X64-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; X64-SLOW-NEXT:    vzeroupper
 ; X64-SLOW-NEXT:    retq
 ;
 ; X64-FAST-LABEL: trunc4:
 ; X64-FAST:       # %bb.0:
 ; X64-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; X64-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
 ; X64-FAST-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0

llvm/trunk/test/CodeGen/X86/avx2-vector-shifts.ll

 entry:
   %shl = lshr <4 x i64> %InVec, <i64 63, i64 63, i64 63, i64 63>
   ret <4 x i64> %shl
 }
 
 define <4 x i32> @srl_trunc_and_v4i64(<4 x i32> %x, <4 x i64> %y) nounwind {
 ; X32-SLOW-LABEL: srl_trunc_and_v4i64:
 ; X32-SLOW:       # %bb.0:
-; X32-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; X32-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; X32-SLOW-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [8,8,8,8]
-; X32-SLOW-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; X32-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; X32-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; X32-SLOW-NEXT:    vbroadcastss {{.*#+}} xmm2 = [8,8,8,8]
+; X32-SLOW-NEXT:    vandps %xmm2, %xmm1, %xmm1
 ; X32-SLOW-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0
 ; X32-SLOW-NEXT:    vzeroupper
 ; X32-SLOW-NEXT:    retl
 ;
 ; X32-FAST-LABEL: srl_trunc_and_v4i64:
 ; X32-FAST:       # %bb.0:
 ; X32-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; X32-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1
 ; X32-FAST-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [8,8,8,8]
 ; X32-FAST-NEXT:    vpand %xmm2, %xmm1, %xmm1
 ; X32-FAST-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0
 ; X32-FAST-NEXT:    vzeroupper
 ; X32-FAST-NEXT:    retl
 ;
 ; X64-SLOW-LABEL: srl_trunc_and_v4i64:
 ; X64-SLOW:       # %bb.0:
-; X64-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; X64-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; X64-SLOW-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [8,8,8,8]
-; X64-SLOW-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; X64-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; X64-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; X64-SLOW-NEXT:    vbroadcastss {{.*#+}} xmm2 = [8,8,8,8]
+; X64-SLOW-NEXT:    vandps %xmm2, %xmm1, %xmm1
 ; X64-SLOW-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0
 ; X64-SLOW-NEXT:    vzeroupper
 ; X64-SLOW-NEXT:    retq
 ;
 ; X64-FAST-LABEL: srl_trunc_and_v4i64:
 ; X64-FAST:       # %bb.0:
 ; X64-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; X64-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1

llvm/trunk/test/CodeGen/X86/combine-shl.ll

 ; SSE41-NEXT:    pslld $23, %xmm1
 ; SSE41-NEXT:    paddd {{.*}}(%rip), %xmm1
 ; SSE41-NEXT:    cvttps2dq %xmm1, %xmm1
 ; SSE41-NEXT:    pmulld %xmm1, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX-SLOW-LABEL: combine_vec_shl_trunc_and:
 ; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX-SLOW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX-SLOW-NEXT:    vandps {{.*}}(%rip), %xmm1, %xmm1
 ; AVX-SLOW-NEXT:    vpsllvd %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    vzeroupper
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: combine_vec_shl_trunc_and:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1

llvm/trunk/test/CodeGen/X86/combine-sra.ll

 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,1,1,1,4,5,6,7]
 ; SSE-NEXT:    psrad %xmm1, %xmm0
 ; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3],xmm0[4,5],xmm5[6,7]
 ; SSE-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: combine_vec_ashr_trunc_and:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vandps {{.*}}(%rip), %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vpsravd %xmm1, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: combine_vec_ashr_trunc_and:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1

 ; SSE-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm0[4,5,6,7]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: combine_vec_ashr_trunc_lshr:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpsravd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: combine_vec_ashr_trunc_lshr:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpsrlq $32, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]

 ; SSE-NEXT:    psrad $1, %xmm2
 ; SSE-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm0[4,5,6,7]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: combine_vec_ashr_trunc_ashr:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
 ; AVX2-SLOW-NEXT:    vpsravd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: combine_vec_ashr_trunc_ashr:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,3,5,7,5,7,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

llvm/trunk/test/CodeGen/X86/combine-srl.ll

 ; SSE-NEXT:    psrld $16, %xmm0
 ; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
 ; SSE-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: combine_vec_lshr_trunc_lshr1:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpsrlvq {{.*}}(%rip), %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: combine_vec_lshr_trunc_lshr1:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpsrlvq {{.*}}(%rip), %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]

 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,1,1,1,4,5,6,7]
 ; SSE-NEXT:    psrld %xmm1, %xmm0
 ; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3],xmm0[4,5],xmm5[6,7]
 ; SSE-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: combine_vec_lshr_trunc_and:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vandps {{.*}}(%rip), %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: combine_vec_lshr_trunc_and:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1

llvm/trunk/test/CodeGen/X86/oddshuffles.ll

 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
 ; AVX1-NEXT:    vmovaps %xmm0, 32(%rdi)
 ; AVX1-NEXT:    vmovaps %ymm2, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: v12i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm3 = ymm0[3,3,2,3,7,7,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1,2],xmm2[3]
-; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm3 = <0,4,u,1,5,u,2,6>
-; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm3, %ymm0
-; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm1
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
-; AVX2-SLOW-NEXT:    vmovaps %ymm0, (%rdi)
-; AVX2-SLOW-NEXT:    vmovaps %xmm2, 32(%rdi)
+; AVX2-SLOW-NEXT:    vmovaps {{.*#+}} ymm2 = <0,4,u,1,5,u,2,6>
+; AVX2-SLOW-NEXT:    vpermps %ymm0, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vbroadcastsd %xmm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3],xmm3[3,3]
+; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3]
+; AVX2-SLOW-NEXT:    vmovaps %xmm0, 32(%rdi)
+; AVX2-SLOW-NEXT:    vmovaps %ymm2, (%rdi)
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: v12i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = <0,4,u,1,5,u,2,6>
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm2
 ; AVX2-FAST-NEXT:    vbroadcastsd %xmm1, %ymm3

llvm/trunk/test/CodeGen/X86/reduce-trunc-shl.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+sse2 | FileCheck %s --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 | FileCheck %s --check-prefix=AVX2
 
 define void @trunc_shl_7_v4i32_v4i64(<4 x i32> addrspace(1)* %out, <4 x i64> addrspace(1)* %in) {
 ; SSE2-LABEL: trunc_shl_7_v4i32_v4i64:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movaps (%rsi), %xmm0
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; SSE2-NEXT:    pslld $7, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, (%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_shl_7_v4i32_v4i64:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vmovaps (%rsi), %xmm0
+; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX2-NEXT:    vpslld $7, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovdqa %xmm0, (%rdi)
-; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
   %val = load <4 x i64>, <4 x i64> addrspace(1)* %in
   %shl = shl <4 x i64> %val, <i64 7, i64 7, i64 7, i64 7>
   %trunc = trunc <4 x i64> %shl to <4 x i32>
   store <4 x i32> %trunc, <4 x i32> addrspace(1)* %out
   ret void
 }
 

llvm/trunk/test/CodeGen/X86/shuffle-vs-trunc-256-widen.ll

 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, (%rsi)
-; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm0 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps (%rdi), %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, (%rsi)
 ; AVX2-FAST-NEXT:    vzeroupper

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i16_with_zext_return_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i16_with_zext_return_v8i16:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i16_via_v4i32_return_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i16_via_v4i32_return_v8i16:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

llvm/trunk/test/CodeGen/X86/shuffle-vs-trunc-256.ll

 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX1-NEXT:    vmovaps %xmm0, (%rsi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX2-SLOW-NEXT:    vmovaps %xmm0, (%rsi)
-; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm0 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps (%rdi), %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps %xmm0, (%rsi)
 ; AVX2-FAST-NEXT:    vzeroupper

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i16_return_v2i64:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i16_return_v2i64:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i16_with_zext_return_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i16_with_zext_return_v8i16:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i16_via_v4i32_return_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i16_via_v4i32_return_v8i16:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i16_return_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13],zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i16_return_v8i16:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[u],zero
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i8_return_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[u],zero
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i8_return_v16i8:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
 ; AVX1-NEXT:    vmovq %xmm0, (%rsi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vmovq %xmm0, (%rsi)
-; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i16:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm0 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd (%rdi), %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
 ; AVX2-FAST-NEXT:    vmovq %xmm0, (%rsi)

 ; AVX1-NEXT:    vmovaps (%rdi), %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
 ; AVX1-NEXT:    vmovd %xmm0, (%rsi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_v4i64_to_v4i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = mem[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vmovaps (%rdi), %xmm0
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],mem[0,2]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
 ; AVX2-SLOW-NEXT:    vmovd %xmm0, (%rsi)
-; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_v4i64_to_v4i8:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm0 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd (%rdi), %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u]
 ; AVX2-FAST-NEXT:    vmovd %xmm0, (%rsi)

llvm/trunk/test/CodeGen/X86/vector-trunc-math-widen.ll

 ; AVX1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpaddq %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpaddq %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1,2,3],xmm4[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpaddq %ymm3, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpaddq %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpaddq %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpaddq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpaddq %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpaddq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vpaddq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpaddq %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vpaddq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpaddd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpaddd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpaddw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpaddb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vpsubq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpsubq %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpsubq %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1,2,3],xmm4[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpsubq %ymm3, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsubq %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpsubq %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpsubq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsubq %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpsubq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vpsubq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsubq %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vpsubq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpsubd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpsubd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsubw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpsubw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsubb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpsubb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm7[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpmulld %xmm7, %xmm3, %xmm3
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm6[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm7, %xmm8
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm7[0,2],xmm8[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm7[0,2]
+; AVX2-SLOW-NEXT:    vpmulld %xmm8, %xmm3, %xmm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm6, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,2],xmm7[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm7[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm6, %xmm2, %xmm2
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm6, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm5[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,2],xmm7[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm5, %xmm1, %xmm1
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm4, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm6, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm3, %xmm3
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vandps %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpand %ymm5, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpand %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpand %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT:    vpand %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vandps %ymm4, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vandps %ymm5, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vandps %ymm6, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vandps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vandps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vxorps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpxor %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vxorps %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vxorps %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpxor %ymm5, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpxor %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpxor %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT:    vpxor %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vxorps %ymm4, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vxorps %ymm5, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vxorps %ymm6, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vxorps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vxorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vxorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpor %ymm3, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpor %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vorps %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vorps %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpor %ymm5, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpor %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpor %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT:    vpor %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vorps %ymm4, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vorps %ymm5, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vorps %ymm6, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vorps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

llvm/trunk/test/CodeGen/X86/vector-trunc-math.ll

 ; AVX1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpaddq %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpaddq %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1,2,3],xmm4[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpaddq %ymm3, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpaddq %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpaddq %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpaddq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpaddq %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpaddq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vpaddq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpaddq %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vpaddq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpaddd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpaddd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpaddw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_add_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_add_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpaddb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vpsubq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpsubq %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpsubq %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1,2,3],xmm4[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpsubq %ymm3, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsubq %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpsubq %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpsubq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsubq %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpsubq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vpsubq %ymm5, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsubq %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vpsubq %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpsubd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpsubd {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsubw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpsubw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_sub_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsubb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_sub_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpsubb {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm7 = ymm7[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm7 = ymm7[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpmulld %xmm7, %xmm3, %xmm3
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm6 = ymm6[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm6 = ymm6[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm7, %xmm8
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm8 = xmm7[0,2],xmm8[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm7[0,2]
+; AVX2-SLOW-NEXT:    vpmulld %xmm8, %xmm3, %xmm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm6, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm6 = xmm6[0,2],xmm7[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm7[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm6, %xmm2, %xmm2
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm6, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm5 = ymm5[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm5 = ymm5[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm5, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,2],xmm7[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm7
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm5, %xmm1, %xmm1
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm4 = ymm4[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm4, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm4[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm6, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_mul_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_mul_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm3, %xmm3
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
 ; AVX2-SLOW-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vandps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vandps %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpand %ymm5, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpand %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpand %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT:    vpand %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vandps %ymm4, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vandps %ymm5, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vandps %ymm6, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vandps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vandps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_and_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_and_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vxorps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpxor %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vxorps %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vxorps %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpxor %ymm5, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpxor %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpxor %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT:    vpxor %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vxorps %ymm4, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vxorps %ymm5, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vxorps %ymm6, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vxorps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vxorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vxorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_xor_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_xor_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vorps %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpor %ymm3, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpor %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vorps %ymm2, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vorps %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpor %ymm5, %ymm1, %ymm1
-; AVX2-SLOW-NEXT:    vpor %ymm4, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpor %ymm7, %ymm3, %ymm3
-; AVX2-SLOW-NEXT:    vpor %ymm6, %ymm2, %ymm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vorps %ymm4, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vorps %ymm5, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vorps %ymm6, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vorps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_v16i64_v16i8:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_const_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vorps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_const_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_const_v8i64_v8i16:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_or_const_v8i64_v8i16:
 ; AVX2-FAST:       # %bb.0:

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_or_const_v16i64_v16i8:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm3 = ymm3[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm5[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpand %xmm4, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

llvm/trunk/test/CodeGen/X86/vector-trunc-packus-widen.ll

 ; AVX2-SLOW-LABEL: trunc_packus_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm1
 ; AVX2-SLOW-NEXT:    vpand %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_packus_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-FAST-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpand %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_packus_v8i64_v8i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm3
-; AVX2-SLOW-NEXT:    vpand %ymm1, %ymm3, %ymm1
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT:    vpand %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vpand %ymm0, %ymm3, %ymm0
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm2
+; AVX2-SLOW-NEXT:    vpand %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_packus_v8i64_v8i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
 ; AVX2-FAST-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm3

llvm/trunk/test/CodeGen/X86/vector-trunc-packus.ll

 ; AVX2-SLOW-LABEL: trunc_packus_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm1
 ; AVX2-SLOW-NEXT:    vpand %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_packus_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-FAST-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vpand %xmm0, %xmm3, %xmm0
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_packus_v8i64_v8i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm3
-; AVX2-SLOW-NEXT:    vpand %ymm1, %ymm3, %ymm1
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT:    vpand %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vpand %ymm0, %ymm3, %ymm0
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm2
+; AVX2-SLOW-NEXT:    vpand %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_packus_v8i64_v8i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
 ; AVX2-FAST-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm3

llvm/trunk/test/CodeGen/X86/vector-trunc-ssat-widen.ll

 ; AVX2-SLOW-LABEL: trunc_ssat_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2147483647,2147483647,2147483647,2147483647]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18446744071562067968,18446744071562067968,18446744071562067968,18446744071562067968]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm2
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_ssat_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2147483647,2147483647,2147483647,2147483647]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-FAST-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_ssat_v8i64_v8i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [2147483647,2147483647,2147483647,2147483647]
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [18446744071562067968,18446744071562067968,18446744071562067968,18446744071562067968]
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm3
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_ssat_v8i64_v8i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [2147483647,2147483647,2147483647,2147483647]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
 ; AVX2-FAST-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0

llvm/trunk/test/CodeGen/X86/vector-trunc-ssat.ll

 ; AVX2-SLOW-LABEL: trunc_ssat_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2147483647,2147483647,2147483647,2147483647]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18446744071562067968,18446744071562067968,18446744071562067968,18446744071562067968]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm2
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_ssat_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2147483647,2147483647,2147483647,2147483647]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm2
 ; AVX2-FAST-NEXT:    vblendvpd %ymm2, %ymm0, %ymm1, %ymm0

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_ssat_v8i64_v8i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [2147483647,2147483647,2147483647,2147483647]
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm3
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [18446744071562067968,18446744071562067968,18446744071562067968,18446744071562067968]
-; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm3
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vpcmpgtq %ymm2, %ymm1, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_ssat_v8i64_v8i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [2147483647,2147483647,2147483647,2147483647]
 ; AVX2-FAST-NEXT:    vpcmpgtq %ymm0, %ymm2, %ymm3
 ; AVX2-FAST-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0

llvm/trunk/test/CodeGen/X86/vector-trunc-usat-widen.ll

 ; AVX2-SLOW-LABEL: trunc_usat_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
 ; AVX2-SLOW-NEXT:    vpxor %ymm1, %ymm0, %ymm1
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [9223372041149743103,9223372041149743103,9223372041149743103,9223372041149743103]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm1
 ; AVX2-SLOW-NEXT:    vmovapd {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,429496729]
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm1, %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_usat_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
 ; AVX2-FAST-NEXT:    vpxor %ymm1, %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [9223372041149743103,9223372041149743103,9223372041149743103,9223372041149743103]

 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_usat_v8i64_v8i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm3 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
-; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm1, %ymm4
+; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm0, %ymm4
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm5 = [9223372041149743103,9223372041149743103,9223372041149743103,9223372041149743103]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm4, %ymm5, %ymm4
-; AVX2-SLOW-NEXT:    vblendvpd %ymm4, %ymm1, %ymm2, %ymm1
-; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm4, %ymm0, %ymm2, %ymm0
+; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm1, %ymm3
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm3, %ymm5, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_usat_v8i64_v8i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm3 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
 ; AVX2-FAST-NEXT:    vpxor %ymm3, %ymm1, %ymm4

llvm/trunk/test/CodeGen/X86/vector-trunc-usat.ll

 ; AVX2-SLOW-LABEL: trunc_usat_v4i64_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
 ; AVX2-SLOW-NEXT:    vpxor %ymm1, %ymm0, %ymm1
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [9223372041149743103,9223372041149743103,9223372041149743103,9223372041149743103]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm1, %ymm2, %ymm1
 ; AVX2-SLOW-NEXT:    vmovapd {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,429496729]
 ; AVX2-SLOW-NEXT:    vblendvpd %ymm1, %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_usat_v4i64_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
 ; AVX2-FAST-NEXT:    vpxor %ymm1, %ymm0, %ymm1
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [9223372041149743103,9223372041149743103,9223372041149743103,9223372041149743103]

 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc_usat_v8i64_v8i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm3 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
-; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm1, %ymm4
+; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm0, %ymm4
 ; AVX2-SLOW-NEXT:    vpbroadcastq {{.*#+}} ymm5 = [9223372041149743103,9223372041149743103,9223372041149743103,9223372041149743103]
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm4, %ymm5, %ymm4
-; AVX2-SLOW-NEXT:    vblendvpd %ymm4, %ymm1, %ymm2, %ymm1
-; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT:    vblendvpd %ymm4, %ymm0, %ymm2, %ymm0
+; AVX2-SLOW-NEXT:    vpxor %ymm3, %ymm1, %ymm3
 ; AVX2-SLOW-NEXT:    vpcmpgtq %ymm3, %ymm5, %ymm3
-; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm0, %ymm2, %ymm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vblendvpd %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc_usat_v8i64_v8i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295]
 ; AVX2-FAST-NEXT:    vpbroadcastq {{.*#+}} ymm3 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
 ; AVX2-FAST-NEXT:    vpxor %ymm3, %ymm1, %ymm4

llvm/trunk/test/CodeGen/X86/vector-trunc-widen.ll

 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i32:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i32:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1

 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm3[1,3]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i32_ashr:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i32_ashr:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [1,3,5,7,5,7,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1

 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm3[1,3]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i32_lshr:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i32_lshr:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vpsrlq $32, %ymm1, %ymm1
 ; AVX2-FAST-NEXT:    vpsrlq $32, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm2, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i16:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i16:
 ; AVX2-FAST:       # %bb.0: # %entry

 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc2x4i64_8i32:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc2x4i64_8i32:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1

llvm/trunk/test/CodeGen/X86/vector-trunc.ll

 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i32:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i32:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1

 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm3[1,3]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i32_ashr:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i32_ashr:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [1,3,5,7,5,7,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1

 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm3[1,3]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i32_lshr:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i32_lshr:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vpsrlq $32, %ymm1, %ymm1
 ; AVX2-FAST-NEXT:    vpsrlq $32, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm2, %ymm0

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4],xmm3[5,6,7]
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i16:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
 ; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
 ; AVX2-SLOW-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i16:
 ; AVX2-FAST:       # %bb.0: # %entry

 ; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackuswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, (%rax)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc8i64_8i8:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u>
 ; AVX2-SLOW-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX2-SLOW-NEXT:    vmovq %xmm0, (%rax)
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;

 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc2x4i64_8i32:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc2x4i64_8i32:
 ; AVX2-FAST:       # %bb.0: # %entry
 ; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
 ; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
 ; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1

 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: trunc2x4i64_8i16:
 ; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; AVX2-SLOW-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
 ; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX2-SLOW-NEXT:    vzeroupper
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc2x4i64_8i16: