Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -8117,6 +8117,32 @@
       return LD;
   }
 
+  // If this is a splat of pairs of 32-bit elements, we can use a narrower
+  // build_vector and broadcast it.
+  // TODO: We could probably generalize this more.
+  if (Subtarget.hasAVX2() && EVTBits == 32 && Values.size() == 2) {
+    SDValue Ops[4] = { Op.getOperand(0), Op.getOperand(1),
+                       DAG.getUNDEF(EltVT), DAG.getUNDEF(EltVT) };
+    auto CanSplat = [](SDValue Op, unsigned NumElems, ArrayRef<SDValue> Ops) {
+      // Make sure all the even/odd operands match.
+      for (unsigned i = 2; i != NumElems; ++i)
+        if (Ops[i % 2] != Op.getOperand(i))
+          return false;
+      return true;
+    };
+    if (CanSplat(Op, NumElems, Ops)) {
+      MVT WideEltVT = VT.isFloatingPoint() ? MVT::f64 : MVT::i64;
+      MVT NarrowVT = MVT::getVectorVT(EltVT, 4);
+      // Create a new build vector and cast to v2i64/v2f64.
+      SDValue NewBV = DAG.getBitcast(MVT::getVectorVT(WideEltVT, 2),
+                                     DAG.getBuildVector(NarrowVT, dl, Ops));
+      // Broadcast from v2i64/v2f64 and cast to final VT.
+      MVT BcastVT = MVT::getVectorVT(WideEltVT, NumElems/2);
+      return DAG.getBitcast(VT, DAG.getNode(X86ISD::VBROADCAST, dl, BcastVT,
+                                            NewBV));
+    }
+  }
+
   // For AVX-length vectors, build the individual 128-bit pieces and use
   // shuffles to put them in place.
   if (VT.getSizeInBits() > 128) {
Index: test/CodeGen/X86/avx2-vbroadcast.ll
===================================================================
--- test/CodeGen/X86/avx2-vbroadcast.ll
+++ test/CodeGen/X86/avx2-vbroadcast.ll
@@ -189,12 +189,7 @@
 ; X32-LABEL: Q64:
 ; X32:       ## %bb.0: ## %entry
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl (%eax), %ecx
-; X32-NEXT:    movl 4(%eax), %eax
-; X32-NEXT:    vmovd %ecx, %xmm0
-; X32-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
+; X32-NEXT:    vpbroadcastq (%eax), %xmm0
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: Q64:
@@ -212,13 +207,8 @@
 ; X32-LABEL: QQ64:
 ; X32:       ## %bb.0: ## %entry
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl (%eax), %ecx
-; X32-NEXT:    movl 4(%eax), %eax
-; X32-NEXT:    vmovd %ecx, %xmm0
-; X32-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
-; X32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: QQ64:
@@ -1380,12 +1370,8 @@
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; X32-NEXT:    vmovaps %xmm0, (%esp)
-; X32-NEXT:    movl (%eax), %ecx
-; X32-NEXT:    movl 4(%eax), %eax
-; X32-NEXT:    vmovd %ecx, %xmm1
-; X32-NEXT:    vpinsrd $1, %eax, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $2, %ecx, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $3, %eax, %xmm1, %xmm1
+; X32-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT:    vpbroadcastq %xmm1, %xmm1
 ; X32-NEXT:    vmovaps %xmm0, {{[0-9]+}}(%esp)
 ; X32-NEXT:    vmovdqa %xmm1, {{[0-9]+}}(%esp)
 ; X32-NEXT:    addl $60, %esp
@@ -1438,15 +1424,10 @@
 ; X32-NEXT:    movl 8(%ebp), %eax
 ; X32-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; X32-NEXT:    vmovaps %ymm0, (%esp)
-; X32-NEXT:    movl (%eax), %ecx
-; X32-NEXT:    movl 4(%eax), %eax
-; X32-NEXT:    vmovd %ecx, %xmm1
-; X32-NEXT:    vpinsrd $1, %eax, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $2, %ecx, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $3, %eax, %xmm1, %xmm1
-; X32-NEXT:    vinserti128 $1, %xmm1, %ymm1, %ymm1
+; X32-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; X32-NEXT:    vbroadcastsd %xmm1, %ymm1
 ; X32-NEXT:    vmovaps %ymm0, {{[0-9]+}}(%esp)
-; X32-NEXT:    vmovdqa %ymm1, {{[0-9]+}}(%esp)
+; X32-NEXT:    vmovaps %ymm1, {{[0-9]+}}(%esp)
 ; X32-NEXT:    movl %ebp, %esp
 ; X32-NEXT:    popl %ebp
 ; X32-NEXT:    vzeroupper
Index: test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
===================================================================
--- test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
+++ test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
@@ -485,16 +485,11 @@
 define <8 x i64> @test_mm512_mask_set1_epi64(<8 x i64> %__O, i8 zeroext %__M, i64 %__A) {
 ; X32-LABEL: test_mm512_mask_set1_epi64:
 ; X32:       # %bb.0: # %entry
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X32-NEXT:    vmovd %edx, %xmm1
-; X32-NEXT:    vpinsrd $1, %ecx, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $2, %edx, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $3, %ecx, %xmm1, %xmm1
-; X32-NEXT:    vinserti128 $1, %xmm1, %ymm1, %ymm1
+; X32-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; X32-NEXT:    vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1
 ; X32-NEXT:    kmovw %eax, %k1
-; X32-NEXT:    vinserti64x4 $1, %ymm1, %zmm1, %zmm0 {%k1}
+; X32-NEXT:    vpbroadcastq %xmm1, %zmm0 {%k1}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_mm512_mask_set1_epi64:
@@ -513,16 +508,11 @@
 define <8 x i64> @test_mm512_maskz_set1_epi64(i8 zeroext %__M, i64 %__A)  {
 ; X32-LABEL: test_mm512_maskz_set1_epi64:
 ; X32:       # %bb.0: # %entry
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X32-NEXT:    vmovd %edx, %xmm0
-; X32-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $3, %ecx, %xmm0, %xmm0
-; X32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X32-NEXT:    vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0
 ; X32-NEXT:    kmovw %eax, %k1
-; X32-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0 {%k1} {z}
+; X32-NEXT:    vpbroadcastq %xmm0, %zmm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_mm512_maskz_set1_epi64:
Index: test/CodeGen/X86/avx512vl-intrinsics-fast-isel.ll
===================================================================
--- test/CodeGen/X86/avx512vl-intrinsics-fast-isel.ll
+++ test/CodeGen/X86/avx512vl-intrinsics-fast-isel.ll
@@ -797,16 +797,11 @@
 define <4 x i64> @test_mm256_mask_set1_epi64(<4 x i64> %__O, i8 zeroext %__M, i64 %__A) {
 ; X32-LABEL: test_mm256_mask_set1_epi64:
 ; X32:       # %bb.0: # %entry
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movb {{[0-9]+}}(%esp), %dl
-; X32-NEXT:    vmovd %ecx, %xmm1
-; X32-NEXT:    vpinsrd $1, %eax, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $2, %ecx, %xmm1, %xmm1
-; X32-NEXT:    vpinsrd $3, %eax, %xmm1, %xmm1
-; X32-NEXT:    vinserti128 $1, %xmm1, %ymm1, %ymm1
-; X32-NEXT:    kmovw %edx, %k1
-; X32-NEXT:    vmovdqa64 %ymm1, %ymm0 {%k1}
+; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X32-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; X32-NEXT:    vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1
+; X32-NEXT:    kmovw %eax, %k1
+; X32-NEXT:    vpbroadcastq %xmm1, %ymm0 {%k1}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_mm256_mask_set1_epi64:
@@ -826,16 +821,11 @@
 define <4 x i64> @test_mm256_maskz_set1_epi64(i8 zeroext %__M, i64 %__A)  {
 ; X32-LABEL: test_mm256_maskz_set1_epi64:
 ; X32:       # %bb.0: # %entry
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movb {{[0-9]+}}(%esp), %dl
-; X32-NEXT:    vmovd %ecx, %xmm0
-; X32-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
-; X32-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
-; X32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
-; X32-NEXT:    kmovw %edx, %k1
-; X32-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; X32-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X32-NEXT:    vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0
+; X32-NEXT:    kmovw %eax, %k1
+; X32-NEXT:    vpbroadcastq %xmm0, %ymm0 {%k1} {z}
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test_mm256_maskz_set1_epi64:
Index: test/CodeGen/X86/broadcastm-lowering.ll
===================================================================
--- test/CodeGen/X86/broadcastm-lowering.ll
+++ test/CodeGen/X86/broadcastm-lowering.ll
@@ -122,9 +122,7 @@
 ; X86-AVX512VLCDBW-NEXT:    kmovd %k0, %eax
 ; X86-AVX512VLCDBW-NEXT:    movzbl %al, %eax
 ; X86-AVX512VLCDBW-NEXT:    vmovd %eax, %xmm0
-; X86-AVX512VLCDBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3],zero,zero,zero,zero,xmm0[0,1,2,3],zero,zero,zero,zero
-; X86-AVX512VLCDBW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
-; X86-AVX512VLCDBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; X86-AVX512VLCDBW-NEXT:    vpbroadcastq %xmm0, %zmm0
 ; X86-AVX512VLCDBW-NEXT:    retl
 entry:
   %0 = icmp eq <8 x i32> %a, %b
@@ -160,8 +158,7 @@
 ; X86-AVX512VLCDBW-NEXT:    kmovd %k0, %eax
 ; X86-AVX512VLCDBW-NEXT:    movzbl %al, %eax
 ; X86-AVX512VLCDBW-NEXT:    vmovd %eax, %xmm0
-; X86-AVX512VLCDBW-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3],zero,zero,zero,zero,xmm0[0,1,2,3],zero,zero,zero,zero
-; X86-AVX512VLCDBW-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX512VLCDBW-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; X86-AVX512VLCDBW-NEXT:    retl
 entry:
   %0 = icmp eq <8 x i32> %a, %b
Index: test/CodeGen/X86/insertelement-shuffle.ll
===================================================================
--- test/CodeGen/X86/insertelement-shuffle.ll
+++ test/CodeGen/X86/insertelement-shuffle.ll
@@ -103,14 +103,9 @@
 ; X32_AVX256-NEXT:    subl $8, %esp
 ; X32_AVX256-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; X32_AVX256-NEXT:    vmovlps %xmm0, (%esp)
-; X32_AVX256-NEXT:    movl (%esp), %eax
-; X32_AVX256-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32_AVX256-NEXT:    vmovd %eax, %xmm0
-; X32_AVX256-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0
-; X32_AVX256-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0
-; X32_AVX256-NEXT:    vpinsrd $3, %ecx, %xmm0, %xmm0
-; X32_AVX256-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
-; X32_AVX256-NEXT:    vmovdqa %ymm0, %ymm1
+; X32_AVX256-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X32_AVX256-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X32_AVX256-NEXT:    vmovaps %ymm0, %ymm1
 ; X32_AVX256-NEXT:    movl %ebp, %esp
 ; X32_AVX256-NEXT:    popl %ebp
 ; X32_AVX256-NEXT:    retl
Index: test/CodeGen/X86/vector-shuffle-combining-xop.ll
===================================================================
--- test/CodeGen/X86/vector-shuffle-combining-xop.ll
+++ test/CodeGen/X86/vector-shuffle-combining-xop.ll
@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefix=X32
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefix=X32
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefix=X64
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefix=X64
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefix=X32 --check-prefix=X86AVX
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefix=X32 --check-prefix=X86AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefix=X64 --check-prefix=X64AVX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefix=X64 --check-prefix=X64AVX2
 
 declare <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double>, <2 x double>, <2 x i64>, i8) nounwind readnone
 declare <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double>, <4 x double>, <4 x i64>, i8) nounwind readnone
@@ -320,20 +320,35 @@
 
 ; FIXME: Duplicated load in i686
 define void @buildvector_v4f32_0404(float %a, float %b, <4 x float>* %ptr) {
-; X32-LABEL: buildvector_v4f32_0404:
-; X32:       # %bb.0:
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; X32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
-; X32-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
-; X32-NEXT:    vmovaps %xmm0, (%eax)
-; X32-NEXT:    retl
+; X86AVX-LABEL: buildvector_v4f32_0404:
+; X86AVX:       # %bb.0:
+; X86AVX-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86AVX-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X86AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
+; X86AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
+; X86AVX-NEXT:    vmovaps %xmm0, (%eax)
+; X86AVX-NEXT:    retl
 ;
-; X64-LABEL: buildvector_v4f32_0404:
-; X64:       # %bb.0:
-; X64-NEXT:    vpermil2ps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[0],xmm1[0]
-; X64-NEXT:    vmovaps %xmm0, (%rdi)
-; X64-NEXT:    retq
+; X86AVX2-LABEL: buildvector_v4f32_0404:
+; X86AVX2:       # %bb.0:
+; X86AVX2-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86AVX2-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X86AVX2-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X86AVX2-NEXT:    vmovapd %xmm0, (%eax)
+; X86AVX2-NEXT:    retl
+;
+; X64AVX-LABEL: buildvector_v4f32_0404:
+; X64AVX:       # %bb.0:
+; X64AVX-NEXT:    vpermil2ps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[0],xmm1[0]
+; X64AVX-NEXT:    vmovaps %xmm0, (%rdi)
+; X64AVX-NEXT:    retq
+;
+; X64AVX2-LABEL: buildvector_v4f32_0404:
+; X64AVX2:       # %bb.0:
+; X64AVX2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
+; X64AVX2-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X64AVX2-NEXT:    vmovapd %xmm0, (%rdi)
+; X64AVX2-NEXT:    retq
   %v0 = insertelement <4 x float> undef, float %a, i32 0
   %v1 = insertelement <4 x float> %v0,   float %b, i32 1
   %v2 = insertelement <4 x float> %v1,   float %a, i32 2