Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -5639,11 +5639,11 @@
       (1 + LastLoadedElt - FirstLoadedElt) * LDBaseVT.getStoreSizeInBits();
 
   // VZEXT_LOAD - consecutive load/undefs followed by zeros/undefs.
-  // TODO: The code below fires only for for loading the low 64-bits of a
-  // of a 128-bit vector. It's probably worth generalizing more.
   if (IsConsecutiveLoad && FirstLoadedElt == 0 && LoadSize == 64 &&
-      (VT.is128BitVector() && TLI.isTypeLegal(MVT::v2i64))) {
-    SDVTList Tys = DAG.getVTList(MVT::v2i64, MVT::Other);
+      ((VT.is128BitVector() && TLI.isTypeLegal(MVT::v2i64)) ||
+       (VT.is256BitVector() && TLI.isTypeLegal(MVT::v4i64)))) {
+    EVT VecVT = VT.is256BitVector() ? MVT::v4i64 : MVT::v2i64;
+    SDVTList Tys = DAG.getVTList(VecVT, MVT::Other);
     SDValue Ops[] = { LDBase->getChain(), LDBase->getBasePtr() };
     SDValue ResNode =
         DAG.getMemIntrinsicNode(X86ISD::VZEXT_LOAD, DL, Tys, Ops, MVT::i64,
Index: lib/Target/X86/X86InstrFragmentsSIMD.td
===================================================================
--- lib/Target/X86/X86InstrFragmentsSIMD.td
+++ lib/Target/X86/X86InstrFragmentsSIMD.td
@@ -682,11 +682,6 @@
   return cast<LoadSDNode>(N)->getAlignment() >= 16;
 }]>;
 
-// Like 'X86vzload', but always requires 128-bit vector alignment.
-def alignedX86vzload : PatFrag<(ops node:$ptr), (X86vzload node:$ptr), [{
-  return cast<MemSDNode>(N)->getAlignment() >= 16;
-}]>;
-
 // Like 'load', but always requires 256-bit vector alignment.
 def alignedload256 : PatFrag<(ops node:$ptr), (load node:$ptr), [{
   return cast<LoadSDNode>(N)->getAlignment() >= 32;
Index: lib/Target/X86/X86InstrSSE.td
===================================================================
--- lib/Target/X86/X86InstrSSE.td
+++ lib/Target/X86/X86InstrSSE.td
@@ -5058,6 +5058,8 @@
   def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
               (v2i64 (scalar_to_vector (loadi64 addr:$src))), (iPTR 0)))),
             (SUBREG_TO_REG (i64 0), (VMOVZQI2PQIrm addr:$src), sub_xmm)>;
+  def : Pat<(v4i64 (X86vzload addr:$src)),
+            (SUBREG_TO_REG (i64 0), (VMOVZQI2PQIrm addr:$src), sub_xmm)>;
 }
 
 let Predicates = [UseSSE2], AddedComplexity = 20 in {
@@ -5066,13 +5068,6 @@
   def : Pat<(v2i64 (X86vzload addr:$src)), (MOVZQI2PQIrm addr:$src)>;
 }
 
-let Predicates = [HasAVX] in {
-def : Pat<(v4i64 (alignedX86vzload addr:$src)),
-          (SUBREG_TO_REG (i32 0), (VMOVAPSrm addr:$src), sub_xmm)>;
-def : Pat<(v4i64 (X86vzload addr:$src)),
-          (SUBREG_TO_REG (i32 0), (VMOVUPSrm addr:$src), sub_xmm)>;
-}
-
 //===---------------------------------------------------------------------===//
 // Moving from XMM to XMM and clear upper 64 bits. Note, there is a bug in
 // IA32 document. movq xmm1, xmm2 does clear the high bits.
Index: test/CodeGen/X86/merge-consecutive-loads-256.ll
===================================================================
--- test/CodeGen/X86/merge-consecutive-loads-256.ll
+++ test/CodeGen/X86/merge-consecutive-loads-256.ll
@@ -214,19 +214,10 @@
 }
 
 define <8 x float> @merge_8f32_f32_12zzuuzz(float* %ptr) nounwind uwtable noinline ssp {
-; AVX1-LABEL: merge_8f32_f32_12zzuuzz:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
-; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: merge_8f32_f32_12zzuuzz:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
+; AVX-LABEL: merge_8f32_f32_12zzuuzz:
+; AVX:       # BB#0:
+; AVX-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; AVX-NEXT:    retq
   %ptr0 = getelementptr inbounds float, float* %ptr, i64 1
   %ptr1 = getelementptr inbounds float, float* %ptr, i64 2
   %val0 = load float, float* %ptr0