Index: llvm/lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- llvm/lib/Target/X86/X86ISelLowering.cpp
+++ llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -40912,7 +40912,52 @@
   return SDValue();
 }
 
+static SDValue combineMovmskPackss(SDNode *Movmsk, SelectionDAG &DAG,
+                                   const X86Subtarget &Subtarget) {
+  // Make sure we can replace a 128-bit MOVMSK with a 256-bit MOVMSK.
+  // TODO: It may be possible to ease the i32/i64 constraint to allow more
+  // optimization with AVX1.
+  MVT MovmskVecVT = Movmsk->getOperand(0).getSimpleValueType();
+  if (!MovmskVecVT.is128BitVector())
+    return SDValue();
+  MVT EltVT = MovmskVecVT.getVectorElementType();
+  bool CanMovmsk256FP =
+      ((EltVT == MVT::f32 || EltVT == MVT::f64) && Subtarget.hasAVX()) ||
+      ((EltVT == MVT::i32 || EltVT == MVT::i64) && Subtarget.hasAVX2());
+  bool CanMovmsk_v32i8 = EltVT == MVT::i8 && Subtarget.hasAVX2();
+  if (!CanMovmsk256FP && !CanMovmsk_v32i8)
+    return SDValue();
+
+  // There has to be a PACKSS operand to this MOVMSK.
+  SDValue Packss = peekThroughBitcasts(Movmsk->getOperand(0));
+  if (Packss.getOpcode() != X86ISD::PACKSS)
+    return SDValue();
+
+  // The PACKSS must use 2 extracted halves of a common source vector.
+  SDValue Extract0 = peekThroughBitcasts(Packss.getOperand(0));
+  SDValue Extract1 = peekThroughBitcasts(Packss.getOperand(1));
+  if (Extract0.getOpcode() != ISD::EXTRACT_SUBVECTOR ||
+      Extract1.getOpcode() != ISD::EXTRACT_SUBVECTOR)
+    return SDValue();
+  SDValue X = Extract0.getOperand(0);
+  MVT XVT = X.getSimpleValueType();
+  if (X != Extract1.getOperand(0) || !XVT.is256BitVector())
+    return SDValue();
+  if (!isNullConstant(Extract0.getOperand(1)))
+    return SDValue();
+  auto *Index1C = dyn_cast<ConstantSDNode>(Extract1.getOperand(1));
+  if (!Index1C || Index1C->getZExtValue() != XVT.getVectorNumElements() / 2)
+    return SDValue();
+
+  // movmsk (packss (extract X, 0), (extract X, 2)) --> movmsk (bitcast X)
+  MVT VT256 = MVT::getVectorVT(EltVT, MovmskVecVT.getVectorNumElements() * 2);
+  SDValue CastX = DAG.getBitcast(VT256, X);
+  MVT VT = Movmsk->getSimpleValueType(0);
+  return DAG.getNode(X86ISD::MOVMSK, SDLoc(Movmsk), VT, CastX);
+}
+
 static SDValue combineMOVMSK(SDNode *N, SelectionDAG &DAG,
+                             const X86Subtarget &Subtarget,
                              TargetLowering::DAGCombinerInfo &DCI) {
   SDValue Src = N->getOperand(0);
   MVT SrcVT = Src.getSimpleValueType();
@@ -40975,6 +41020,9 @@
     }
   }
 
+  if (SDValue V = combineMovmskPackss(N, DAG, Subtarget))
+    return V;
+
   return SDValue();
 }
 
@@ -42771,7 +42819,7 @@
   case X86ISD::FMSUBADD_RND:
   case X86ISD::FMADDSUB:
   case X86ISD::FMSUBADD:    return combineFMADDSUB(N, DAG, Subtarget);
-  case X86ISD::MOVMSK:      return combineMOVMSK(N, DAG, DCI);
+  case X86ISD::MOVMSK:      return combineMOVMSK(N, DAG, Subtarget, DCI);
   case X86ISD::MGATHER:
   case X86ISD::MSCATTER:
   case ISD::MGATHER:
Index: llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll
===================================================================
--- llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll
+++ llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll
@@ -76,9 +76,7 @@
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpgtq %ymm3, %ymm2, %ymm1
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovmskps %xmm0, %eax
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -121,17 +119,27 @@
 ; SSE2-SSSE3-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX12-LABEL: v4f64:
-; AVX12:       # %bb.0:
-; AVX12-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX12-NEXT:    vcmpltpd %ymm2, %ymm3, %ymm1
-; AVX12-NEXT:    vandpd %ymm1, %ymm0, %ymm0
-; AVX12-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX12-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vmovmskps %xmm0, %eax
-; AVX12-NEXT:    # kill: def $al killed $al killed $eax
-; AVX12-NEXT:    vzeroupper
-; AVX12-NEXT:    retq
+; AVX1-LABEL: v4f64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX1-NEXT:    vcmpltpd %ymm2, %ymm3, %ymm1
+; AVX1-NEXT:    vandpd %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovmskps %xmm0, %eax
+; AVX1-NEXT:    # kill: def $al killed $al killed $eax
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: v4f64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vcmpltpd %ymm2, %ymm3, %ymm1
+; AVX2-NEXT:    vandpd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    # kill: def $al killed $al killed $eax
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: v4f64:
 ; AVX512F:       # %bb.0:
@@ -194,9 +202,7 @@
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpgtw %ymm3, %ymm2, %ymm1
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
Index: llvm/test/CodeGen/X86/bitcast-setcc-256.ll
===================================================================
--- llvm/test/CodeGen/X86/bitcast-setcc-256.ll
+++ llvm/test/CodeGen/X86/bitcast-setcc-256.ll
@@ -31,9 +31,7 @@
 ; AVX2-LABEL: v16i16:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -384,9 +382,7 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    movw %ax, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
Index: llvm/test/CodeGen/X86/bitcast-setcc-512.ll
===================================================================
--- llvm/test/CodeGen/X86/bitcast-setcc-512.ll
+++ llvm/test/CodeGen/X86/bitcast-setcc-512.ll
@@ -109,13 +109,10 @@
 ;
 ; AVX2-LABEL: v16i32:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpgtd %ymm3, %ymm1, %ymm1
 ; AVX2-NEXT:    vpcmpgtd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -170,13 +167,10 @@
 ;
 ; AVX2-LABEL: v16f32:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vcmpltps %ymm1, %ymm3, %ymm1
 ; AVX2-NEXT:    vcmpltps %ymm0, %ymm2, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    # kill: def $ax killed $ax killed $eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
@@ -592,14 +586,11 @@
 ;
 ; AVX2-LABEL: bitcast_16i32_store:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm2, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    movw %ax, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
Index: llvm/test/CodeGen/X86/movmsk-cmp.ll
===================================================================
--- llvm/test/CodeGen/X86/movmsk-cmp.ll
+++ llvm/test/CodeGen/X86/movmsk-cmp.ll
@@ -433,9 +433,7 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    cmpw $-1, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -495,9 +493,7 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    testw %ax, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -905,14 +901,11 @@
 ;
 ; AVX2-LABEL: allones_v16i32_sign:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm2, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    cmpw $-1, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -979,14 +972,11 @@
 ;
 ; AVX2-LABEL: allzeros_v16i32_sign:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm2, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    testw %ax, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -1856,9 +1846,7 @@
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    cmpw $-1, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -2121,9 +2109,7 @@
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    testw %ax, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -2411,16 +2397,12 @@
 ;
 ; AVX2-LABEL: allones_v16i32_and1:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1]
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    cmpw $-1, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -2496,16 +2478,12 @@
 ;
 ; AVX2-LABEL: allzeros_v16i32_and1:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1]
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    testw %ax, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -3455,9 +3433,7 @@
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    cmpw $-1, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -3720,9 +3696,7 @@
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    testw %ax, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -4010,16 +3984,12 @@
 ;
 ; AVX2-LABEL: allones_v16i32_and4:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [4,4,4,4,4,4,4,4]
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    cmpw $-1, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
@@ -4095,16 +4065,12 @@
 ;
 ; AVX2-LABEL: allzeros_v16i32_and4:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [4,4,4,4,4,4,4,4]
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackssdw %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    testw %ax, %ax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
Index: llvm/test/CodeGen/X86/vector-compare-all_of.ll
===================================================================
--- llvm/test/CodeGen/X86/vector-compare-all_of.ll
+++ llvm/test/CodeGen/X86/vector-compare-all_of.ll
@@ -91,18 +91,29 @@
 ; SSE-NEXT:    cmovneq %rcx, %rax
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: test_v4f64_legal_sext:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovmskps %xmm0, %eax
-; AVX-NEXT:    xorl %ecx, %ecx
-; AVX-NEXT:    cmpl $15, %eax
-; AVX-NEXT:    movq $-1, %rax
-; AVX-NEXT:    cmovneq %rcx, %rax
-; AVX-NEXT:    vzeroupper
-; AVX-NEXT:    retq
+; AVX1-LABEL: test_v4f64_legal_sext:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovmskps %xmm0, %eax
+; AVX1-NEXT:    xorl %ecx, %ecx
+; AVX1-NEXT:    cmpl $15, %eax
+; AVX1-NEXT:    movq $-1, %rax
+; AVX1-NEXT:    cmovneq %rcx, %rax
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: test_v4f64_legal_sext:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl $15, %eax
+; AVX2-NEXT:    movq $-1, %rax
+; AVX2-NEXT:    cmovneq %rcx, %rax
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v4f64_legal_sext:
 ; AVX512:       # %bb.0:
@@ -229,18 +240,29 @@
 ; SSE-NEXT:    cmovnel %ecx, %eax
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: test_v8f32_legal_sext:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpmovmskb %xmm0, %eax
-; AVX-NEXT:    xorl %ecx, %ecx
-; AVX-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
-; AVX-NEXT:    movl $-1, %eax
-; AVX-NEXT:    cmovnel %ecx, %eax
-; AVX-NEXT:    vzeroupper
-; AVX-NEXT:    retq
+; AVX1-LABEL: test_v8f32_legal_sext:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpmovmskb %xmm0, %eax
+; AVX1-NEXT:    xorl %ecx, %ecx
+; AVX1-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX1-NEXT:    movl $-1, %eax
+; AVX1-NEXT:    cmovnel %ecx, %eax
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: test_v8f32_legal_sext:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX2-NEXT:    movl $-1, %eax
+; AVX2-NEXT:    cmovnel %ecx, %eax
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v8f32_legal_sext:
 ; AVX512:       # %bb.0:
@@ -389,9 +411,7 @@
 ; AVX2-LABEL: test_v4i64_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovmskps %xmm0, %eax
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
 ; AVX2-NEXT:    xorl %ecx, %ecx
 ; AVX2-NEXT:    cmpl $15, %eax
 ; AVX2-NEXT:    movq $-1, %rax
@@ -557,9 +577,7 @@
 ; AVX2-LABEL: test_v8i32_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    xorl %ecx, %ecx
 ; AVX2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
 ; AVX2-NEXT:    movl $-1, %eax
@@ -746,9 +764,7 @@
 ; AVX2-LABEL: test_v16i16_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    xorl %ecx, %ecx
 ; AVX2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
 ; AVX2-NEXT:    movl $-1, %eax
@@ -1506,9 +1522,7 @@
 ; AVX2-LABEL: bool_reduction_v16i16:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    xorl %ecx, %ecx
 ; AVX2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
 ; AVX2-NEXT:    movl $-1, %eax
Index: llvm/test/CodeGen/X86/vector-compare-any_of.ll
===================================================================
--- llvm/test/CodeGen/X86/vector-compare-any_of.ll
+++ llvm/test/CodeGen/X86/vector-compare-any_of.ll
@@ -90,18 +90,29 @@
 ; SSE-NEXT:    movslq %ecx, %rax
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: test_v4f64_legal_sext:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovmskps %xmm0, %eax
-; AVX-NEXT:    xorl %ecx, %ecx
-; AVX-NEXT:    cmpl %eax, %ecx
-; AVX-NEXT:    sbbl %ecx, %ecx
-; AVX-NEXT:    movslq %ecx, %rax
-; AVX-NEXT:    vzeroupper
-; AVX-NEXT:    retq
+; AVX1-LABEL: test_v4f64_legal_sext:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovmskps %xmm0, %eax
+; AVX1-NEXT:    xorl %ecx, %ecx
+; AVX1-NEXT:    cmpl %eax, %ecx
+; AVX1-NEXT:    sbbl %ecx, %ecx
+; AVX1-NEXT:    movslq %ecx, %rax
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: test_v4f64_legal_sext:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    xorl %ecx, %ecx
+; AVX2-NEXT:    cmpl %eax, %ecx
+; AVX2-NEXT:    sbbl %ecx, %ecx
+; AVX2-NEXT:    movslq %ecx, %rax
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v4f64_legal_sext:
 ; AVX512:       # %bb.0:
@@ -223,17 +234,27 @@
 ; SSE-NEXT:    sbbl %eax, %eax
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: test_v8f32_legal_sext:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpmovmskb %xmm0, %ecx
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    cmpl %ecx, %eax
-; AVX-NEXT:    sbbl %eax, %eax
-; AVX-NEXT:    vzeroupper
-; AVX-NEXT:    retq
+; AVX1-LABEL: test_v8f32_legal_sext:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX1-NEXT:    xorl %eax, %eax
+; AVX1-NEXT:    cmpl %ecx, %eax
+; AVX1-NEXT:    sbbl %eax, %eax
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: test_v8f32_legal_sext:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpmovmskb %ymm0, %ecx
+; AVX2-NEXT:    xorl %eax, %eax
+; AVX2-NEXT:    cmpl %ecx, %eax
+; AVX2-NEXT:    sbbl %eax, %eax
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test_v8f32_legal_sext:
 ; AVX512:       # %bb.0:
@@ -380,9 +401,7 @@
 ; AVX2-LABEL: test_v4i64_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovmskps %xmm0, %eax
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
 ; AVX2-NEXT:    xorl %ecx, %ecx
 ; AVX2-NEXT:    cmpl %eax, %ecx
 ; AVX2-NEXT:    sbbl %ecx, %ecx
@@ -541,9 +560,7 @@
 ; AVX2-LABEL: test_v8i32_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX2-NEXT:    vpmovmskb %ymm0, %ecx
 ; AVX2-NEXT:    xorl %eax, %eax
 ; AVX2-NEXT:    cmpl %ecx, %eax
 ; AVX2-NEXT:    sbbl %eax, %eax
@@ -724,9 +741,7 @@
 ; AVX2-LABEL: test_v16i16_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX2-NEXT:    vpmovmskb %ymm0, %ecx
 ; AVX2-NEXT:    xorl %eax, %eax
 ; AVX2-NEXT:    cmpl %ecx, %eax
 ; AVX2-NEXT:    sbbl %eax, %eax
@@ -1494,9 +1509,7 @@
 ; AVX2-LABEL: bool_reduction_v16i16:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX2-NEXT:    vpmovmskb %ymm0, %ecx
 ; AVX2-NEXT:    xorl %eax, %eax
 ; AVX2-NEXT:    cmpl %ecx, %eax
 ; AVX2-NEXT:    sbbl %eax, %eax