[X86][SSE] Improve i16 splatting shuffles

[RKSimon]

Summary

Better handling of the annoying pshuflw/pshufhw ops which only shuffle lower/upper halves of a vector.

Added vXi16 unary shuffle support for cases where i16 elements (from the same half of the source) are being splatted to the whole of one of the halves. This avoids the general lowering case which must shuffle the 32-bit elements first - meaning that we used to end up with unnecessary duplicate pshuflw/pshufhw shuffles.

Note this has the side effect of a lot of SSSE3 test cases no longer needing to use PSHUFB, as it falls below the 3 op combine threshold for when PSHUFB is typically worth it.

[RKSimon]

Ping - adding extra subscribers.

[RKSimon]

ping?

[RKSimon]

ping*2

[qcolombet]

test/CodeGen/X86/avx-splat.ll

18

I am missing something.

You said:

Note this has the side effect of a lot of SSSE3 test cases no longer needing to use PSHUFB, as it falls below the 3 op combine threshold for when PSHUFB is typically worth it.

But isn’t pshufb always better than two other pshufX instructions?

[RKSimon]

test/CodeGen/X86/avx-splat.ll

18

On all but the newest machines (Haswell+) PSHUFB has a 3 or more cy latency (which is where the 3 op threshold comes from). The bigger problem is that PSHUFB nearly always has to load the shuffle mask operand from the constant pool, which is a lot more costly than 3 immediate ops.

[escha]

Just a side note, but Agner claims pshufb is 1 cycle latency on Wolfdale, Nehalem, and Ivy Bridge as well.

[RKSimon]

In D14901#312358, @escha wrote:

Just a side note, but Agner claims pshufb is 1 cycle latency on Wolfdale, Nehalem, and Ivy Bridge as well.

But not any recent Atom or AMD targets - it still doesn't account for the cost of loading the shuffle mask either unfortunately. On the whole I think the 3 op threshold is about right.

[RKSimon]

Tested on Jaguar CPU:

Throughput:
Old 3op shuffle: 4cy
New 2op shuffle: 2cy
pshufb_rr 3cy
pshufb_rm 3cy

Latency:
Old 3op shuffle: 4cy
New 2op shuffle: 3cy
pshufb_rr 3cy
pshufb_rm 4cy

[qcolombet]

Thanks for the numbers Simon.

[RKSimon]

Sorry Quentin - I missed your follow up email to the list - copied here:

Tested on Jaguar CPU:

Throughput:
Old 3op shuffle: 4cy
New 2op shuffle: 2cy
pshufb_rr 3cy
pshufb_rm 3cy

I am confused.
When the code sequence is shorter, I was expecting this, but this number is not for the problem we were discussing, i.e., when the shufb is replaced by 2 shuf(w|hd, whatever), right?
If it is, I am missing something because it should be 2 uops in both cases.

I'm confused too - I'm not certain what outstanding problem with my patch you think I should be addressing.

What it does is improve vXi16 shuffles so that more patterns can be performed in 2uops instead of 3uops, a side effect of which is that a later combine stage in PerformShuffleCombine (combineX86ShufflesRecursively) no longer merges these into a single PSHUFB as its threshold for combining is 3uops. The timing tests I did demonstrated that this threshold is probably about right - although I accept that more recent targets can perform PSHUFB faster.

What am I missing?

[qcolombet]

Hi Simon,

[RKSimon]

I've raised a bugzilla: https://llvm.org/bugs/show_bug.cgi?id=26183 to discuss what the threshold for combines to PSHUFB should be and how it should be controlled.

[qcolombet]

Hi Simon,

Thanks for filing a PR.

You’re right the threshold problem is separate, LGTM then.

Cheers,
-Quentin

lib/Target/X86/X86ISelLowering.cpp

static SDValue lowerV8I16GeneralSingleInputVectorShuffle(
   int NumLToH =
       std::lower_bound(HiInputs.begin(), HiInputs.end(), 4) - HiInputs.begin();
   int NumHToH = HiInputs.size() - NumLToH;
   MutableArrayRef<int> LToLInputs(LoInputs.data(), NumLToL);
   MutableArrayRef<int> LToHInputs(HiInputs.data(), NumLToH);
   MutableArrayRef<int> HToLInputs(LoInputs.data() + NumLToL, NumHToL);
   MutableArrayRef<int> HToHInputs(HiInputs.data() + NumLToH, NumHToH);
 
+  // If we are splatting two values from one half - one to each half, then
+  // we can shuffle that half so each is splatted to a dword, then splat those
+  // to their respective halves.
+  auto SplatHalfs = [&](int LoInput, int HiInput, unsigned ShufWOp,
+                        int DOffset) {
+    int PSHUFHalfMask[] = {LoInput % 4, LoInput % 4, HiInput % 4, HiInput % 4};
+    int PSHUFDMask[] = {DOffset + 0, DOffset + 0, DOffset + 1, DOffset + 1};
+    V = DAG.getNode(ShufWOp, DL, VT, V,
+                    getV4X86ShuffleImm8ForMask(PSHUFHalfMask, DL, DAG));
+    V = DAG.getBitcast(PSHUFDVT, V);
+    V = DAG.getNode(X86ISD::PSHUFD, DL, PSHUFDVT, V,
+                    getV4X86ShuffleImm8ForMask(PSHUFDMask, DL, DAG));
+    return DAG.getBitcast(VT, V);
+  };
+
+  if (NumLToL == 1 && NumLToH == 1 && (NumHToL + NumHToH) == 0)
+    return SplatHalfs(LToLInputs[0], LToHInputs[0], X86ISD::PSHUFLW, 0);
+  if (NumHToL == 1 && NumHToH == 1 && (NumLToL + NumLToH) == 0)
+    return SplatHalfs(HToLInputs[0], HToHInputs[0], X86ISD::PSHUFHW, 2);
+
   // Simplify the 1-into-3 and 3-into-1 cases with a single pshufd. For all
   // such inputs we can swap two of the dwords across the half mark and end up
   // with <=2 inputs to each half in each half. Once there, we can fall through
   // to the generic code below. For example:
   //
   // Input: [a, b, c, d, e, f, g, h] -PSHUFD[0,2,1,3]-> [a, b, e, f, c, d, g, h]
   // Mask:  [0, 1, 2, 7, 4, 5, 6, 3] -----------------> [0, 1, 4, 7, 2, 3, 6, 5]
   //

test/CodeGen/X86/avx-splat.ll

 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx | FileCheck %s
 
 define <32 x i8> @funcA(<32 x i8> %a) nounwind uwtable readnone ssp {
 ; CHECK-LABEL: funcA:
 ; CHECK:       ## BB#0: ## %entry
 ; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
 ; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; CHECK-NEXT:    retq
 entry:
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> undef, <32 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <32 x i8> %shuffle
 }
 
 define <16 x i16> @funcB(<16 x i16> %a) nounwind uwtable readnone ssp {
 ; CHECK-LABEL: funcB:
 ; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[10,11,10,11,10,11,10,11,10,11,10,11,10,11,10,11]
+; CHECK-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; CHECK-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; CHECK-NEXT:    retq
 entry:
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> undef, <16 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
   ret <16 x i16> %shuffle
 }
 
 define <4 x i64> @funcC(i64 %q) nounwind uwtable readnone ssp {

test/CodeGen/X86/psubus.ll

vector.ph:
   %3 = icmp ugt <8 x i16> %2, <i16 32766, i16 32766, i16 32766, i16 32766, i16 32766, i16 32766, i16 32766, i16 32766>
   %4 = add <8 x i16> %2, <i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767>
   %5 = select <8 x i1> %3, <8 x i16> %4, <8 x i16> zeroinitializer
   store <8 x i16> %5, <8 x i16>* %1, align 2
   ret void
 }
 
 define void @test3(i16* nocapture %head, i16 zeroext %w) nounwind {
-; SSE2-LABEL: test3:
-; SSE2:       ## BB#0: ## %vector.ph
-; SSE2-NEXT:    movd %esi, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE2-NEXT:    movdqu (%rdi), %xmm1
-; SSE2-NEXT:    psubusw %xmm0, %xmm1
-; SSE2-NEXT:    movdqu %xmm1, (%rdi)
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: test3:
-; SSSE3:       ## BB#0: ## %vector.ph
-; SSSE3-NEXT:    movd %esi, %xmm0
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSSE3-NEXT:    movdqu (%rdi), %xmm1
-; SSSE3-NEXT:    psubusw %xmm0, %xmm1
-; SSSE3-NEXT:    movdqu %xmm1, (%rdi)
-; SSSE3-NEXT:    retq
+; SSE-LABEL: test3:
+; SSE:       ## BB#0: ## %vector.ph
+; SSE-NEXT:    movd %esi, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqu (%rdi), %xmm1
+; SSE-NEXT:    psubusw %xmm0, %xmm1
+; SSE-NEXT:    movdqu %xmm1, (%rdi)
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test3:
 ; AVX1:       ## BB#0: ## %vector.ph
 ; AVX1-NEXT:    vmovd %esi, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vmovdqu (%rdi), %xmm1
 ; AVX1-NEXT:    vpsubusw %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test3:
 ; AVX2:       ## BB#0: ## %vector.ph
 ; AVX2-NEXT:    vmovd %esi, %xmm0
vector.ph:
   ret void
 }
 
 define void @test6(i8* nocapture %head, i8 zeroext %w) nounwind {
 ; SSE2-LABEL: test6:
 ; SSE2:       ## BB#0: ## %vector.ph
 ; SSE2-NEXT:    movd %esi, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    movdqu (%rdi), %xmm1
 ; SSE2-NEXT:    psubusb %xmm0, %xmm1
 ; SSE2-NEXT:    movdqu %xmm1, (%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: test6:
 ; SSSE3:       ## BB#0: ## %vector.ph
 ; SSSE3-NEXT:    movd %esi, %xmm0
vector.ph:
   %4 = add <16 x i16> %2, <i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767, i16 -32767>
   %5 = select <16 x i1> %3, <16 x i16> %4, <16 x i16> zeroinitializer
   store <16 x i16> %5, <16 x i16>* %1, align 2
   ret void
 
 }
 
 define void @test9(i16* nocapture %head, i16 zeroext %w) nounwind {
-; SSE2-LABEL: test9:
-; SSE2:       ## BB#0: ## %vector.ph
-; SSE2-NEXT:    movd %esi, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE2-NEXT:    movdqu (%rdi), %xmm1
-; SSE2-NEXT:    movdqu 16(%rdi), %xmm2
-; SSE2-NEXT:    psubusw %xmm0, %xmm1
-; SSE2-NEXT:    psubusw %xmm0, %xmm2
-; SSE2-NEXT:    movdqu %xmm2, 16(%rdi)
-; SSE2-NEXT:    movdqu %xmm1, (%rdi)
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: test9:
-; SSSE3:       ## BB#0: ## %vector.ph
-; SSSE3-NEXT:    movd %esi, %xmm0
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSSE3-NEXT:    movdqu (%rdi), %xmm1
-; SSSE3-NEXT:    movdqu 16(%rdi), %xmm2
-; SSSE3-NEXT:    psubusw %xmm0, %xmm1
-; SSSE3-NEXT:    psubusw %xmm0, %xmm2
-; SSSE3-NEXT:    movdqu %xmm2, 16(%rdi)
-; SSSE3-NEXT:    movdqu %xmm1, (%rdi)
-; SSSE3-NEXT:    retq
+; SSE-LABEL: test9:
+; SSE:       ## BB#0: ## %vector.ph
+; SSE-NEXT:    movd %esi, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    movdqu (%rdi), %xmm1
+; SSE-NEXT:    movdqu 16(%rdi), %xmm2
+; SSE-NEXT:    psubusw %xmm0, %xmm1
+; SSE-NEXT:    psubusw %xmm0, %xmm2
+; SSE-NEXT:    movdqu %xmm2, 16(%rdi)
+; SSE-NEXT:    movdqu %xmm1, (%rdi)
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test9:
 ; AVX1:       ## BB#0: ## %vector.ph
 ; AVX1-NEXT:    vmovups (%rdi), %ymm0
-; AVX1-NEXT:    vmovd %esi, %xmm1
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vpsubw %xmm1, %xmm2, %xmm3
-; AVX1-NEXT:    vpsubw %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vmovd %esi, %xmm2
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; AVX1-NEXT:    vpsubw %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpsubw %xmm2, %xmm0, %xmm4
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm4, %ymm3
-; AVX1-NEXT:    vpmaxuw %xmm1, %xmm2, %xmm4
-; AVX1-NEXT:    vpcmpeqw %xmm4, %xmm2, %xmm2
-; AVX1-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm1
-; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpmaxuw %xmm2, %xmm1, %xmm4
+; AVX1-NEXT:    vpcmpeqw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmaxuw %xmm2, %xmm0, %xmm2
+; AVX1-NEXT:    vpcmpeqw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vandps %ymm3, %ymm0, %ymm0
 ; AVX1-NEXT:    vmovups %ymm0, (%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test9:
 ; AVX2:       ## BB#0: ## %vector.ph
 ; AVX2-NEXT:    vmovd %esi, %xmm0
vector.ph:
   ret void
 }
 
 define void @test12(i8* nocapture %head, i8 zeroext %w) nounwind {
 ; SSE2-LABEL: test12:
 ; SSE2:       ## BB#0: ## %vector.ph
 ; SSE2-NEXT:    movd %esi, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    movdqu (%rdi), %xmm1
 ; SSE2-NEXT:    movdqu 16(%rdi), %xmm2
 ; SSE2-NEXT:    psubusb %xmm0, %xmm1
 ; SSE2-NEXT:    psubusb %xmm0, %xmm2
 ; SSE2-NEXT:    movdqu %xmm2, 16(%rdi)
 ; SSE2-NEXT:    movdqu %xmm1, (%rdi)
 ; SSE2-NEXT:    retq
 ;

test/CodeGen/X86/vector-shift-ashr-128.ll

; X32-SSE-NEXT:    retl
   %shift = ashr <8 x i16> %a, %splat
   ret <8 x i16> %shift
 }
 
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, <16 x i8> %b) nounwind {
 ; SSE2-LABEL: splatvar_shift_v16i8:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
 ; SSE2-NEXT:    psllw $5, %xmm3
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
 ; SSE2-NEXT:    pxor %xmm5, %xmm5
 ; SSE2-NEXT:    pcmpgtw %xmm4, %xmm5
 ; SSE2-NEXT:    movdqa %xmm5, %xmm6
 ; SSE2-NEXT:    pandn %xmm1, %xmm6

 ; XOPAVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX2-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
 ; XOPAVX2-NEXT:    vpshab %xmm1, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    retq
 ;
 ; X32-SSE-LABEL: splatvar_shift_v16i8:
 ; X32-SSE:       # BB#0:
 ; X32-SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; X32-SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; X32-SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,4,4]
+; X32-SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
 ; X32-SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
 ; X32-SSE-NEXT:    psllw $5, %xmm3
 ; X32-SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
 ; X32-SSE-NEXT:    pxor %xmm2, %xmm2
 ; X32-SSE-NEXT:    pxor %xmm5, %xmm5
 ; X32-SSE-NEXT:    pcmpgtw %xmm4, %xmm5
 ; X32-SSE-NEXT:    movdqa %xmm5, %xmm6
 ; X32-SSE-NEXT:    pandn %xmm1, %xmm6

test/CodeGen/X86/vector-shift-lshr-128.ll

; X32-SSE-NEXT:    retl
   %shift = lshr <8 x i16> %a, %splat
   ret <8 x i16> %shift
 }
 
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, <16 x i8> %b) nounwind {
 ; SSE2-LABEL: splatvar_shift_v16i8:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
 ; SSE2-NEXT:    psllw $5, %xmm2
 ; SSE2-NEXT:    pxor %xmm1, %xmm1
 ; SSE2-NEXT:    pxor %xmm3, %xmm3
 ; SSE2-NEXT:    pcmpgtb %xmm2, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    pandn %xmm0, %xmm4
 ; SSE2-NEXT:    psrlw $4, %xmm0
 ; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0

 ; XOPAVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; XOPAVX2-NEXT:    vpsubb %xmm1, %xmm2, %xmm1
 ; XOPAVX2-NEXT:    vpshlb %xmm1, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    retq
 ;
 ; X32-SSE-LABEL: splatvar_shift_v16i8:
 ; X32-SSE:       # BB#0:
 ; X32-SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; X32-SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; X32-SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,4,4,4]
+; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
 ; X32-SSE-NEXT:    psllw $5, %xmm2
 ; X32-SSE-NEXT:    pxor %xmm1, %xmm1
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    pcmpgtb %xmm2, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm3, %xmm4
 ; X32-SSE-NEXT:    pandn %xmm0, %xmm4
 ; X32-SSE-NEXT:    psrlw $4, %xmm0
 ; X32-SSE-NEXT:    pand .LCPI7_0, %xmm0

test/CodeGen/X86/vector-shift-shl-128.ll

; X32-SSE-NEXT:    retl
   %shift = shl <8 x i16> %a, %splat
   ret <8 x i16> %shift
 }
 
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, <16 x i8> %b) nounwind {
 ; SSE2-LABEL: splatvar_shift_v16i8:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
 ; SSE2-NEXT:    psllw $5, %xmm2
 ; SSE2-NEXT:    pxor %xmm1, %xmm1
 ; SSE2-NEXT:    pxor %xmm3, %xmm3
 ; SSE2-NEXT:    pcmpgtb %xmm2, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    pandn %xmm0, %xmm4
 ; SSE2-NEXT:    psllw $4, %xmm0
 ; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0

 ; XOPAVX2:       # BB#0:
 ; XOPAVX2-NEXT:    vpbroadcastb %xmm1, %xmm1
 ; XOPAVX2-NEXT:    vpshlb %xmm1, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    retq
 ;
 ; X32-SSE-LABEL: splatvar_shift_v16i8:
 ; X32-SSE:       # BB#0:
 ; X32-SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; X32-SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; X32-SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,4,4,4]
+; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
 ; X32-SSE-NEXT:    psllw $5, %xmm2
 ; X32-SSE-NEXT:    pxor %xmm1, %xmm1
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    pcmpgtb %xmm2, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm3, %xmm4
 ; X32-SSE-NEXT:    pandn %xmm0, %xmm4
 ; X32-SSE-NEXT:    psllw $4, %xmm0
 ; X32-SSE-NEXT:    pand .LCPI7_0, %xmm0

test/CodeGen/X86/vector-shuffle-128-v16.ll

 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-unknown"
 
 define <16 x i8> @shuffle_v16i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i8> %a, <16 x i8> %b) {
-; FIXME: SSE2 should look like the following:
-; FIXME-LABEL: @shuffle_v16i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00
-; FIXME:       # BB#0:
-; FIXME-NEXT:    punpcklbw %xmm0, %xmm0
-; FIXME-NEXT:    pshuflw {{.*}} # xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; FIXME-NEXT:    pshufd {{.*}} # xmm0 = xmm0[0,1,0,1]
-; FIXME-NEXT:    retq
-;
 ; SSE2-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    pxor %xmm1, %xmm1
 ; SSSE3-NEXT:    pshufb %xmm1, %xmm0
 ; SSSE3-NEXT:    retq
 ;
; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_00_00_00_00_00_00_00_00_01_01_01_01_01_01_01_01(<16 x i8> %a, <16 x i8> %b) {
 ; SSE2-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_01_01_01_01_01_01_01_01:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,1,1,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_01_01_01_01_01_01_01_01:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_01_01_01_01_01_01_01_01:
 ; SSE41:       # BB#0:
 ; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1]
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_01_01_01_01_01_01_01_01:
 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08(<16 x i8> %a, <16 x i8> %b) {
 ; SSE2-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,2,4,5,6,7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,2,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,8,8,8,8,8,8,8,8]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v16i8_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:

 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 0, i32 1, i32 1, i32 2, i32 2, i32 3, i32 3, i32 4, i32 4, i32 5, i32 5, i32 6, i32 6, i32 7, i32 7>
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_0101010101010101(<16 x i8> %a, <16 x i8> %b) {
-; FIXME: SSE2 should be the following:
-; FIXME-LABEL: @shuffle_v16i8_0101010101010101
-; FIXME:       # BB#0:
-; FIXME-NEXT:    pshuflw {{.*}} # xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; FIXME-NEXT:    pshufd {{.*}} # xmm0 = xmm0[0,1,0,1]
-; FIXME-NEXT:    retq
-;
-; SSE2-LABEL: shuffle_v16i8_0101010101010101:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: shuffle_v16i8_0101010101010101:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: shuffle_v16i8_0101010101010101:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSE41-NEXT:    retq
+; SSE-LABEL: shuffle_v16i8_0101010101010101:
+; SSE:       # BB#0:
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: shuffle_v16i8_0101010101010101:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i8_0101010101010101:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1, i32 0, i32 1>
   ret <16 x i8> %shuffle
; AVX-NEXT:    retq
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_16_00_16_01_16_02_16_03_16_04_16_05_16_06_16_07(<16 x i8> %a, <16 x i8> %b) {
 ; SSE2-LABEL: shuffle_v16i8_16_00_16_01_16_02_16_03_16_04_16_05_16_06_16_07:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE2-NEXT:    pand %xmm2, %xmm1
 ; SSE2-NEXT:    pandn %xmm0, %xmm2
 ; SSE2-NEXT:    por %xmm1, %xmm2
 ; SSE2-NEXT:    movdqa %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v16i8_16_00_16_01_16_02_16_03_16_04_16_05_16_06_16_07:
 ; SSSE3:       # BB#0:

 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,1,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,4]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE2-NEXT:    packuswb %xmm5, %xmm2
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,0,255,255,255,0,255,255,255,0,255,255,255]
 ; SSE2-NEXT:    pand %xmm0, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,3,1,1,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE2-NEXT:    pandn %xmm1, %xmm0
 ; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v16i8_uu_10_02_07_22_14_07_02_18_03_01_14_18_09_11_00:
 ; SSSE3:       # BB#0: # %entry
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 = xmm1[u],zero,zero,zero,xmm1[6],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[2],zero,zero,zero
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[u,10,2,7],zero,xmm0[14,7,2],zero,xmm0[3,1,14],zero,xmm0[9,11,0]
; AVX-NEXT:    retq
   ret <16 x i8> %bitcast8
 }
 
 define <16 x i8> @insert_dup_mem_v16i8_i32(i32* %ptr) {
 ; SSE2-LABEL: insert_dup_mem_v16i8_i32:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_mem_v16i8_i32:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSSE3-NEXT:    pxor %xmm1, %xmm1
 ; SSSE3-NEXT:    pshufb %xmm1, %xmm0
 ; SSSE3-NEXT:    retq

 }
 
 define <16 x i8> @insert_dup_mem_v16i8_sext_i8(i8* %ptr) {
 ; SSE2-LABEL: insert_dup_mem_v16i8_sext_i8:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movsbl (%rdi), %eax
 ; SSE2-NEXT:    movd %eax, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_mem_v16i8_sext_i8:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movsbl (%rdi), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm0
 ; SSSE3-NEXT:    pxor %xmm1, %xmm1
 ; SSSE3-NEXT:    pshufb %xmm1, %xmm0
; AVX2-NEXT:    retq
   ret <16 x i8> %tmp4
 }
 
 define <16 x i8> @insert_dup_elt1_mem_v16i8_i32(i32* %ptr) {
 ; SSE2-LABEL: insert_dup_elt1_mem_v16i8_i32:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_elt1_mem_v16i8_i32:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
 ; SSSE3-NEXT:    retq
 ;
; AVX2-NEXT:    retq
   ret <16 x i8> %tmp3
 }
 
 define <16 x i8> @insert_dup_elt2_mem_v16i8_i32(i32* %ptr) {
 ; SSE2-LABEL: insert_dup_elt2_mem_v16i8_i32:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_elt2_mem_v16i8_i32:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
 ; SSSE3-NEXT:    retq
 ;

 }
 
 define <16 x i8> @insert_dup_elt1_mem_v16i8_sext_i8(i8* %ptr) {
 ; SSE2-LABEL: insert_dup_elt1_mem_v16i8_sext_i8:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movsbl (%rdi), %eax
 ; SSE2-NEXT:    movd %eax, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_elt1_mem_v16i8_sext_i8:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movsbl (%rdi), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm0
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
 ; SSSE3-NEXT:    retq

 }
 
 define <16 x i8> @insert_dup_elt2_mem_v16i8_sext_i8(i8* %ptr) {
 ; SSE2-LABEL: insert_dup_elt2_mem_v16i8_sext_i8:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movsbl (%rdi), %eax
 ; SSE2-NEXT:    movd %eax, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_elt2_mem_v16i8_sext_i8:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movsbl (%rdi), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm0
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
 ; SSSE3-NEXT:    retq

test/CodeGen/X86/vector-shuffle-128-v8.ll

 ; AVX:       # BB#0:
 ; AVX-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_00000000(<8 x i16> %a, <8 x i16> %b) {
-; SSE2-LABEL: shuffle_v8i16_00000000:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: shuffle_v8i16_00000000:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: shuffle_v8i16_00000000:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSE41-NEXT:    retq
+; SSE-LABEL: shuffle_v8i16_00000000:
+; SSE:       # BB#0:
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: shuffle_v8i16_00000000:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v8i16_00000000:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <8 x i16> %shuffle

 ; AVX-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> zeroinitializer, <8 x i32> <i32 8, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 undef>
 
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @insert_dup_mem_v8i16_i32(i32* %ptr) {
-; SSE2-LABEL: insert_dup_mem_v8i16_i32:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: insert_dup_mem_v8i16_i32:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: insert_dup_mem_v8i16_i32:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSE41-NEXT:    retq
+; SSE-LABEL: insert_dup_mem_v8i16_i32:
+; SSE:       # BB#0:
+; SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: insert_dup_mem_v8i16_i32:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_mem_v8i16_i32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw (%rdi), %xmm0
 ; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
   %tmp2 = bitcast <4 x i32> %tmp1 to <8 x i16>
   %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> undef, <8 x i32> zeroinitializer
   ret <8 x i16> %tmp3
 }
 
 define <8 x i16> @insert_dup_mem_v8i16_sext_i16(i16* %ptr) {
-; SSE2-LABEL: insert_dup_mem_v8i16_sext_i16:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    movswl (%rdi), %eax
-; SSE2-NEXT:    movd %eax, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: insert_dup_mem_v8i16_sext_i16:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    movswl (%rdi), %eax
-; SSSE3-NEXT:    movd %eax, %xmm0
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: insert_dup_mem_v8i16_sext_i16:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    movswl (%rdi), %eax
-; SSE41-NEXT:    movd %eax, %xmm0
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; SSE41-NEXT:    retq
+; SSE-LABEL: insert_dup_mem_v8i16_sext_i16:
+; SSE:       # BB#0:
+; SSE-NEXT:    movswl (%rdi), %eax
+; SSE-NEXT:    movd %eax, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: insert_dup_mem_v8i16_sext_i16:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    movswl (%rdi), %eax
 ; AVX1-NEXT:    vmovd %eax, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_mem_v8i16_sext_i16:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    movswl (%rdi), %eax
 ; AVX2-NEXT:    vmovd %eax, %xmm0
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm0
 ; AVX2-NEXT:    retq
   %tmp = load i16, i16* %ptr, align 2
   %tmp1 = sext i16 %tmp to i32
   %tmp2 = insertelement <4 x i32> zeroinitializer, i32 %tmp1, i32 0
   %tmp3 = bitcast <4 x i32> %tmp2 to <8 x i16>
   %tmp4 = shufflevector <8 x i16> %tmp3, <8 x i16> undef, <8 x i32> zeroinitializer
   ret <8 x i16> %tmp4
 }
 
 define <8 x i16> @insert_dup_elt1_mem_v8i16_i32(i32* %ptr) {
-; SSE2-LABEL: insert_dup_elt1_mem_v8i16_i32:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: insert_dup_elt1_mem_v8i16_i32:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: insert_dup_elt1_mem_v8i16_i32:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
-; SSE41-NEXT:    retq
+; SSE-LABEL: insert_dup_elt1_mem_v8i16_i32:
+; SSE:       # BB#0:
+; SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: insert_dup_elt1_mem_v8i16_i32:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_elt1_mem_v8i16_i32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw 2(%rdi), %xmm0
 ; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
   %tmp2 = bitcast <4 x i32> %tmp1 to <8 x i16>
   %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <8 x i16> %tmp3
 }
 
 define <8 x i16> @insert_dup_elt3_mem_v8i16_i32(i32* %ptr) {
 ; SSE2-LABEL: insert_dup_elt3_mem_v8i16_i32:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,1,0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,1,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_elt3_mem_v8i16_i32:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
 ; SSSE3-NEXT:    retq
 ;
; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 1
   %tmp2 = bitcast <4 x i32> %tmp1 to <8 x i16>
   %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
   ret <8 x i16> %tmp3
 }
 
 define <8 x i16> @insert_dup_elt1_mem_v8i16_sext_i16(i16* %ptr) {
-; SSE2-LABEL: insert_dup_elt1_mem_v8i16_sext_i16:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    movswl (%rdi), %eax
-; SSE2-NEXT:    movd %eax, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: insert_dup_elt1_mem_v8i16_sext_i16:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    movswl (%rdi), %eax
-; SSSE3-NEXT:    movd %eax, %xmm0
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: insert_dup_elt1_mem_v8i16_sext_i16:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    movswl (%rdi), %eax
-; SSE41-NEXT:    movd %eax, %xmm0
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
-; SSE41-NEXT:    retq
+; SSE-LABEL: insert_dup_elt1_mem_v8i16_sext_i16:
+; SSE:       # BB#0:
+; SSE-NEXT:    movswl (%rdi), %eax
+; SSE-NEXT:    movd %eax, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: insert_dup_elt1_mem_v8i16_sext_i16:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    movswl (%rdi), %eax
 ; AVX1-NEXT:    vmovd %eax, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_elt1_mem_v8i16_sext_i16:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    movswl (%rdi), %eax
 ; AVX2-NEXT:    shrl $16, %eax
 ; AVX2-NEXT:    vmovd %eax, %xmm0
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm0
 ; AVX2-NEXT:    retq
   %tmp = load i16, i16* %ptr, align 2
   %tmp1 = sext i16 %tmp to i32
   %tmp2 = insertelement <4 x i32> zeroinitializer, i32 %tmp1, i32 0
   %tmp3 = bitcast <4 x i32> %tmp2 to <8 x i16>
   %tmp4 = shufflevector <8 x i16> %tmp3, <8 x i16> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <8 x i16> %tmp4
 }
 
 define <8 x i16> @insert_dup_elt3_mem_v8i16_sext_i16(i16* %ptr) {
 ; SSE2-LABEL: insert_dup_elt3_mem_v8i16_sext_i16:
 ; SSE2:       # BB#0:
 ; SSE2-NEXT:    movswl (%rdi), %eax
 ; SSE2-NEXT:    movd %eax, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,1,0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,1,1]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: insert_dup_elt3_mem_v8i16_sext_i16:
 ; SSSE3:       # BB#0:
 ; SSSE3-NEXT:    movswl (%rdi), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm0
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
 ; SSSE3-NEXT:    retq

test/CodeGen/X86/vector-shuffle-256-v16.ll

 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX1
 ; RUN: llc < %s -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX2
 
 target triple = "x86_64-unknown-unknown"
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,4,4,4]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,4]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,2,3,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_01_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,2,3,0,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_00_02_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_00_03_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,8,9,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_00_04_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,8,9,0,1,0,1,0,1,0,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,10,11,0,1,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_00_05_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,10,11,0,1,0,1,0,1,0,1,0,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,12,13,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_00_06_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,12,13,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[14,15,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_07_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm1
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[14,15,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,0,1,2,3]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,0,1,1,4,4,5,5]
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
 ; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,6,7,0,1]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <0,0,255,255,u,u,u,u,u,u,u,u,u,u,u,u,255,255,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
 ; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,2,3,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 9, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,10,11,0,1,0,1]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 10, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,14,15,0,1,0,1,0,1]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4,5,6,7]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 11, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm1[4,5,6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,8,9,0,1,0,1,0,1,0,1]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_12_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,8,9,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 12, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm1[4,5,6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,10,11,0,1,0,1,0,1,0,1,0,1]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_13_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,10,11,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 13, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm1[4,5,6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,12,13,0,1,0,1,0,1,0,1,0,1,0,1]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_14_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,12,13,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 14, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm1[4,5,6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[14,15,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_15_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[14,15,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [14,15,14,15,14,15,14,15,14,15,14,15,14,15,14,15]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,7,7,7,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[14,15,14,15,14,15,14,15,14,15,14,15,14,15,14,15,30,31,30,31,30,31,30,31,30,31,30,31,30,31,30,31]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,2,3,3,6,6,7,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
 ; AVX1:       # BB#0:

 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,0,1,0,1,0,1,12,13,0,1,16,17,16,17,20,21,16,17,16,17,16,17,28,29,16,17]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,0,1,1,4,4,5,5]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 16, i32 0, i32 16, i32 0, i32 16, i32 0, i32 16, i32 8, i32 24, i32 8, i32 24, i32 8, i32 24, i32 8, i32 24>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_16_16_16_16_04_05_06_07_24_24_24_24_12_13_14_15(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_16_16_16_16_04_05_06_07_24_24_24_24_12_13_14_15:

 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 4, i32 4, i32 4, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_uu_uu_00_00_00_00_00_08_08_uu_uu_08_08_14_08(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_uu_uu_00_00_00_00_00_08_08_uu_uu_08_08_14_08:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,4,5,6,7,0,1,0,1,12,13,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_uu_uu_00_00_00_00_00_08_08_uu_uu_08_08_14_08:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,u,u,u,u,0,1,0,1,0,1,0,1,0,1,16,17,16,17,u,u,u,u,16,17,16,17,28,29,16,17]

 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,2,3]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,2,3]

 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 23, i32 undef, i32 3, i32 undef, i32 20, i32 20, i32 5, i32 undef, i32 31, i32 undef, i32 11, i32 undef, i32 28, i32 28, i32 13, i32 undef>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[6,7,6,7,6,7,6,7,6,7,6,7,6,7,6,7]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_u_u_u_u_u_u_u_u_3_3_3_3_3_3_3_3:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[6,7,6,7,6,7,6,7,6,7,6,7,6,7,6,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8_8:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm0
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
+; AVX2-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @insert_v16i16_0elt_into_zero_vector(i16* %ptr) {
 ; ALL-LABEL: insert_v16i16_0elt_into_zero_vector:
 ; ALL:       # BB#0:
; ALL-NEXT:    retq
   %shuffle16 = bitcast <32 x i8> %shuffle8 to <16 x i16>
   ret <16 x i16> %shuffle16
 }
 
 define <16 x i16> @insert_dup_mem_v16i16_i32(i32* %ptr) {
 ; AVX1-LABEL: insert_dup_mem_v16i16_i32:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_mem_v16i16_i32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw (%rdi), %ymm0
 ; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
   %tmp2 = bitcast <4 x i32> %tmp1 to <8 x i16>
   %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> undef, <16 x i32> zeroinitializer
   ret <16 x i16> %tmp3
 }
 
 define <16 x i16> @insert_dup_mem_v16i16_sext_i16(i16* %ptr) {
 ; AVX1-LABEL: insert_dup_mem_v16i16_sext_i16:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    movswl (%rdi), %eax
 ; AVX1-NEXT:    vmovd %eax, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_mem_v16i16_sext_i16:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    movswl (%rdi), %eax
 ; AVX2-NEXT:    vmovd %eax, %xmm0
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
 ; AVX2-NEXT:    retq
   %tmp = load i16, i16* %ptr, align 2
   %tmp1 = sext i16 %tmp to i32
   %tmp2 = insertelement <4 x i32> zeroinitializer, i32 %tmp1, i32 0
   %tmp3 = bitcast <4 x i32> %tmp2 to <8 x i16>
   %tmp4 = shufflevector <8 x i16> %tmp3, <8 x i16> undef, <16 x i32> zeroinitializer
   ret <16 x i16> %tmp4
 }
 
 define <16 x i16> @insert_dup_elt1_mem_v16i16_i32(i32* %ptr) #0 {
 ; AVX1-LABEL: insert_dup_elt1_mem_v16i16_i32:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_elt1_mem_v16i16_i32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw 2(%rdi), %ymm0
 ; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 0
   %tmp2 = bitcast <4 x i32> %tmp1 to <8 x i16>
   %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> undef, <16 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <16 x i16> %tmp3
 }
 
 define <16 x i16> @insert_dup_elt3_mem_v16i16_i32(i32* %ptr) #0 {
 ; AVX1-LABEL: insert_dup_elt3_mem_v16i16_i32:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,2,3,2,3,2,3,2,3,2,3,2,3,2,3]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,1,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: insert_dup_elt3_mem_v16i16_i32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw 2(%rdi), %ymm0
 ; AVX2-NEXT:    retq
   %tmp = load i32, i32* %ptr, align 4
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %tmp, i32 1
   %tmp2 = bitcast <4 x i32> %tmp1 to <8 x i16>
   %tmp3 = shufflevector <8 x i16> %tmp2, <8 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
   ret <16 x i16> %tmp3
 }

test/CodeGen/X86/vector-shuffle-256-v32.ll

; AVX2-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> zeroinitializer, <32 x i32> <i32 32, i32 1, i32 34, i32 3, i32 36, i32 5, i32 38, i32 7, i32 40, i32 9, i32 42, i32 11, i32 44, i32 13, i32 46, i32 15, i32 48, i32 17, i32 50, i32 19, i32 52, i32 21, i32 54, i32 23, i32 56, i32 25, i32 58, i32 27, i32 60, i32 29, i32 62, i32 31>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
 ; AVX2-NEXT:    vpbroadcastb %xmm1, %ymm1
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
 ; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_16_48_16_48_16_48_16_48_16_48_16_48_16_48_16_48(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_16_48_16_48_16_48_16_48_16_48_16_48_16_48_16_48:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
-; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; AVX1-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_16_48_16_48_16_48_16_48_16_48_16_48_16_48_16_48:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpxor %ymm2, %ymm2, %ymm2
 ; AVX2-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
 ; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 0, i32 32, i32 16, i32 48, i32 16, i32 48, i32 16, i32 48, i32 16, i32 48, i32 16, i32 48, i32 16, i32 48, i32 16, i32 48, i32 16, i32 48>
   ret <32 x i8> %shuffle
 }
 
 define <32 x i8> @shuffle_v32i8_32_32_32_32_32_32_32_32_08_09_10_11_12_13_14_15_48_48_48_48_48_48_48_48_24_25_26_27_28_29_30_31(<32 x i8> %a, <32 x i8> %b) {

 }
 
 define <32 x i8> @shuffle_v32i8_08_08_08_08_08_08_08_08_uu_uu_uu_uu_uu_uu_uu_uu_16_16_16_uu_uu_uu_uu_uu_uu_uu_24_24_24_24_24_24(<32 x i8> %a, <32 x i8> %b) {
 ; AVX1-LABEL: shuffle_v32i8_08_08_08_08_08_08_08_08_uu_uu_uu_uu_uu_uu_uu_uu_16_16_16_uu_uu_uu_uu_uu_uu_uu_24_24_24_24_24_24:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,8,8,9,9,8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,8,8,8,8,8,8,8,8]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v32i8_08_08_08_08_08_08_08_08_uu_uu_uu_uu_uu_uu_uu_uu_16_16_16_uu_uu_uu_uu_uu_uu_uu_24_24_24_24_24_24:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,8,8,8,8,8,8,8,u,u,u,u,u,u,u,u,16,16,16,u,u,u,u,u,u,u,24,24,24,24,24,24]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 16, i32 16, i32 16, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24>