Index: include/llvm/Target/TargetLowering.h
===================================================================
--- include/llvm/Target/TargetLowering.h
+++ include/llvm/Target/TargetLowering.h
@@ -2851,6 +2851,11 @@
   /// \returns MERGE_VALUEs of the scalar loads with their chains.
   SDValue scalarizeVectorLoad(LoadSDNode *LD, SelectionDAG &DAG) const;
 
+  // Turn a store of a vector type into stores of the individual elements.
+  /// \param ST Store with a vector value type
+  /// \returns MERGE_VALUs of the individual store chains.
+  SDValue scalarizeVectorStore(StoreSDNode *ST, SelectionDAG &DAG) const;
+
   //===--------------------------------------------------------------------===//
   // Instruction Emitting Hooks
   //
Index: lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -326,6 +326,12 @@
       ST->getMemoryVT().isVector()) {
     EVT intVT = EVT::getIntegerVT(*DAG.getContext(), VT.getSizeInBits());
     if (TLI.isTypeLegal(intVT)) {
+      if (!TLI.isOperationLegalOrCustom(ISD::STORE, intVT)) {
+        // Scalarize the store and let the individual components be handled.
+        SDValue Result = TLI.scalarizeVectorStore(ST, DAG);
+        DAGLegalize->ReplaceNode(SDValue(ST, 0), Result);
+        return;
+      }
       // Expand to a bitconvert of the value to the integer type of the
       // same size, then a (misaligned) int store.
       // FIXME: Does not handle truncating floating point stores!
Index: lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
+++ lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
@@ -636,54 +636,40 @@
 }
 
 SDValue VectorLegalizer::ExpandStore(SDValue Op) {
-  SDLoc dl(Op);
   StoreSDNode *ST = cast<StoreSDNode>(Op.getNode());
-  SDValue Chain = ST->getChain();
-  SDValue BasePTR = ST->getBasePtr();
-  SDValue Value = ST->getValue();
-  EVT StVT = ST->getMemoryVT();
-
-  unsigned Alignment = ST->getAlignment();
-  bool isVolatile = ST->isVolatile();
-  bool isNonTemporal = ST->isNonTemporal();
-  AAMDNodes AAInfo = ST->getAAInfo();
 
-  unsigned NumElem = StVT.getVectorNumElements();
-  // The type of the data we want to save
-  EVT RegVT = Value.getValueType();
-  EVT RegSclVT = RegVT.getScalarType();
-  // The type of data as saved in memory.
+  EVT StVT = ST->getMemoryVT();
   EVT MemSclVT = StVT.getScalarType();
-
-  // Cast floats into integers
   unsigned ScalarSize = MemSclVT.getSizeInBits();
 
   // Round odd types to the next pow of two.
-  if (!isPowerOf2_32(ScalarSize))
-    ScalarSize = NextPowerOf2(ScalarSize);
-
-  // Store Stride in bytes
-  unsigned Stride = ScalarSize/8;
-  // Extract each of the elements from the original vector
-  // and save them into memory individually.
-  SmallVector<SDValue, 8> Stores;
-  for (unsigned Idx = 0; Idx < NumElem; Idx++) {
-    SDValue Ex = DAG.getNode(
-        ISD::EXTRACT_VECTOR_ELT, dl, RegSclVT, Value,
-        DAG.getConstant(Idx, dl, TLI.getVectorIdxTy(DAG.getDataLayout())));
-
-    // This scalar TruncStore may be illegal, but we legalize it later.
-    SDValue Store = DAG.getTruncStore(Chain, dl, Ex, BasePTR,
-               ST->getPointerInfo().getWithOffset(Idx*Stride), MemSclVT,
-               isVolatile, isNonTemporal, MinAlign(Alignment, Idx*Stride),
-               AAInfo);
-
-    BasePTR = DAG.getNode(ISD::ADD, dl, BasePTR.getValueType(), BasePTR,
-                          DAG.getConstant(Stride, dl, BasePTR.getValueType()));
-
-    Stores.push_back(Store);
+  if (!isPowerOf2_32(ScalarSize)) {
+    // FIXME: This is completely broken and inconsistent with ExpandLoad
+    // handling.
+
+    // For sub-byte element sizes, this ends up with 0 stride between elements,
+    // so the same element just gets re-written to the same location. There seem
+    // to be tests explicitly testing for this broken behavior though.  tests
+    // for this broken behavior.
+
+    LLVMContext &Ctx = *DAG.getContext();
+
+    EVT NewMemVT
+      = EVT::getVectorVT(Ctx,
+                         MemSclVT.getIntegerVT(Ctx, NextPowerOf2(ScalarSize)),
+                         StVT.getVectorNumElements());
+
+    SDValue NewVectorStore
+      = DAG.getTruncStore(ST->getChain(), SDLoc(Op), ST->getValue(),
+                          ST->getBasePtr(),
+                          ST->getPointerInfo(), NewMemVT,
+                          ST->isVolatile(), ST->isNonTemporal(),
+                          ST->getAlignment(),
+                          ST->getAAInfo());
+    ST = cast<StoreSDNode>(NewVectorStore.getNode());
   }
-  SDValue TF =  DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Stores);
+
+  SDValue TF = TLI.scalarizeVectorStore(ST, DAG);
   AddLegalizedOperand(Op, TF);
   return TF;
 }
Index: lib/CodeGen/SelectionDAG/TargetLowering.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -3149,6 +3149,60 @@
   return DAG.getMergeValues({ Value, NewChain }, SL);
 }
 
+// FIXME: This relies on each element having a byte size, otherwise the stride
+// is 0 and just overwrites the same location. ExpandStore currently expects
+// this broken behavior.
+SDValue TargetLowering::scalarizeVectorStore(StoreSDNode *ST,
+                                             SelectionDAG &DAG) const {
+  SDLoc SL(ST);
+
+  SDValue Chain = ST->getChain();
+  SDValue BasePtr = ST->getBasePtr();
+  SDValue Value = ST->getValue();
+  EVT StVT = ST->getMemoryVT();
+
+  unsigned Alignment = ST->getAlignment();
+  bool isVolatile = ST->isVolatile();
+  bool isNonTemporal = ST->isNonTemporal();
+  AAMDNodes AAInfo = ST->getAAInfo();
+
+  // The type of the data we want to save
+  EVT RegVT = Value.getValueType();
+  EVT RegSclVT = RegVT.getScalarType();
+
+  // The type of data as saved in memory.
+  EVT MemSclVT = StVT.getScalarType();
+
+  EVT PtrVT = BasePtr.getValueType();
+
+  // Store Stride in bytes
+  unsigned Stride = MemSclVT.getSizeInBits() / 8;
+  EVT IdxVT = getVectorIdxTy(DAG.getDataLayout());
+  unsigned NumElem = StVT.getVectorNumElements();
+
+  // Extract each of the elements from the original vector and save them into
+  // memory individually.
+  SmallVector<SDValue, 8> Stores;
+  for (unsigned Idx = 0; Idx < NumElem; ++Idx) {
+    SDValue Elt = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SL, RegSclVT, Value,
+                              DAG.getConstant(Idx, SL, IdxVT));
+
+    SDValue Ptr = DAG.getNode(ISD::ADD, SL, PtrVT, BasePtr,
+                              DAG.getConstant(Idx * Stride, SL, PtrVT));
+
+    // This scalar TruncStore may be illegal, but we legalize it later.
+    SDValue Store = DAG.getTruncStore(
+      Chain, SL, Elt, Ptr,
+      ST->getPointerInfo().getWithOffset(Idx * Stride), MemSclVT,
+      isVolatile, isNonTemporal, MinAlign(Alignment, Idx * Stride),
+      AAInfo);
+
+    Stores.push_back(Store);
+  }
+
+  return DAG.getNode(ISD::TokenFactor, SL, MVT::Other, Stores);
+}
+
 //===----------------------------------------------------------------------===//
 // Implementation of Emulated TLS Model
 //===----------------------------------------------------------------------===//