* Implement the ByOuterInner accessors

* use them (big simplification in Assign.h)
* axe (Inner|Outer)StrideAtCompileTime that were just introduced
* ei_int_if_dynamic now asserts that the size is the expected one: adapt to that in Block.h
* add rowStride() / colStride() in DenseBase
* implement innerStride() / outerStride() everywhere needed

22 files changed, 222 insertions, 276 deletions

diff --git a/Eigen/src/Core/Assign.h b/Eigen/src/Core/Assign.h
index 174fd0080..3133aa03a 100644
--- a/Eigen/src/Core/Assign.h
+++ b/Eigen/src/Core/Assign.h
@@ -55,7 +55,9 @@ private:
   };
 
   enum {
-    StorageOrdersAgree = (int(Derived::Flags)&RowMajorBit)==(int(OtherDerived::Flags)&RowMajorBit),
+    LhsIsEffectivelyRowMajor = (Derived::RowsAtCompileTime==1) || (int(Derived::Flags)&RowMajorBit),
+    RhsIsEffectivelyRowMajor = (OtherDerived::RowsAtCompileTime==1) || (int(OtherDerived::Flags)&RowMajorBit),
+    StorageOrdersAgree = (LhsIsEffectivelyRowMajor == RhsIsEffectivelyRowMajor),
     MightVectorize = StorageOrdersAgree
                   && (int(Derived::Flags) & int(OtherDerived::Flags) & ActualPacketAccessBit),
     MayInnerVectorize  = MightVectorize && int(InnerSize)!=Dynamic && int(InnerSize)%int(PacketSize)==0
@@ -139,17 +141,13 @@ template<typename Derived1, typename Derived2, int Index, int Stop>
 struct ei_assign_DefaultTraversal_CompleteUnrolling
 {
   enum {
-    row = int(Derived1::Flags)&RowMajorBit
-        ? Index / int(Derived1::ColsAtCompileTime)
-        : Index % Derived1::RowsAtCompileTime,
-    col = int(Derived1::Flags)&RowMajorBit
-        ? Index % int(Derived1::ColsAtCompileTime)
-        : Index / Derived1::RowsAtCompileTime
+    outer = Index / Derived1::InnerSizeAtCompileTime,
+    inner = Index % Derived1::InnerSizeAtCompileTime
   };
 
   EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src)
   {
-    dst.copyCoeff(row, col, src);
+    dst.copyCoeffByOuterInner(outer, inner, src);
     ei_assign_DefaultTraversal_CompleteUnrolling<Derived1, Derived2, Index+1, Stop>::run(dst, src);
   }
 };
@@ -163,13 +161,10 @@ struct ei_assign_DefaultTraversal_CompleteUnrolling<Derived1, Derived2, Stop, St
 template<typename Derived1, typename Derived2, int Index, int Stop>
 struct ei_assign_DefaultTraversal_InnerUnrolling
 {
-  EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src, int row_or_col)
+  EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src, int outer)
   {
-    const bool rowMajor = int(Derived1::Flags)&RowMajorBit;
-    const int row = rowMajor ? row_or_col : Index;
-    const int col = rowMajor ? Index : row_or_col;
-    dst.copyCoeff(row, col, src);
-    ei_assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, Index+1, Stop>::run(dst, src, row_or_col);
+    dst.copyCoeffByOuterInner(outer, Index, src);
+    ei_assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, Index+1, Stop>::run(dst, src, outer);
   }
 };
 
@@ -207,18 +202,14 @@ template<typename Derived1, typename Derived2, int Index, int Stop>
 struct ei_assign_innervec_CompleteUnrolling
 {
   enum {
-    row = int(Derived1::Flags)&RowMajorBit
-        ? Index / int(Derived1::ColsAtCompileTime)
-        : Index % Derived1::RowsAtCompileTime,
-    col = int(Derived1::Flags)&RowMajorBit
-        ? Index % int(Derived1::ColsAtCompileTime)
-        : Index / Derived1::RowsAtCompileTime,
+    outer = Index / Derived1::InnerSizeAtCompileTime,
+    inner = Index % Derived1::InnerSizeAtCompileTime,
     JointAlignment = ei_assign_traits<Derived1,Derived2>::JointAlignment
   };
 
   EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src)
   {
-    dst.template copyPacket<Derived2, Aligned, JointAlignment>(row, col, src);
+    dst.template copyPacketByOuterInner<Derived2, Aligned, JointAlignment>(outer, inner, src);
     ei_assign_innervec_CompleteUnrolling<Derived1, Derived2,
       Index+ei_packet_traits<typename Derived1::Scalar>::size, Stop>::run(dst, src);
   }
@@ -233,13 +224,11 @@ struct ei_assign_innervec_CompleteUnrolling<Derived1, Derived2, Stop, Stop>
 template<typename Derived1, typename Derived2, int Index, int Stop>
 struct ei_assign_innervec_InnerUnrolling
 {
-  EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src, int row_or_col)
+  EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src, int outer)
   {
-    const int row = int(Derived1::Flags)&RowMajorBit ? row_or_col : Index;
-    const int col = int(Derived1::Flags)&RowMajorBit ? Index : row_or_col;
-    dst.template copyPacket<Derived2, Aligned, Aligned>(row, col, src);
+    dst.template copyPacketByOuterInner<Derived2, Aligned, Aligned>(outer, Index, src);
     ei_assign_innervec_InnerUnrolling<Derived1, Derived2,
-      Index+ei_packet_traits<typename Derived1::Scalar>::size, Stop>::run(dst, src, row_or_col);
+      Index+ei_packet_traits<typename Derived1::Scalar>::size, Stop>::run(dst, src, outer);
   }
 };
 
@@ -267,29 +256,11 @@ struct ei_assign_impl<Derived1, Derived2, DefaultTraversal, NoUnrolling>
 {
   inline static void run(Derived1 &dst, const Derived2 &src)
   {
-    if(Derived1::ColsAtCompileTime == 1)
-    {
-      for(int i = 0; i < dst.rows(); ++i)
-        dst.copyCoeff(i, 0, src);
-    }
-    else if(Derived1::RowsAtCompileTime == 1)
-    {
-      for(int i = 0; i < dst.cols(); ++i)
-        dst.copyCoeff(0, i, src);
-    }
-    else
-    {
-      const int innerSize = dst.innerSize();
-      const int outerSize = dst.outerSize();
-      for(int j = 0; j < outerSize; ++j)
-        for(int i = 0; i < innerSize; ++i)
-        {
-          if(int(Derived1::Flags)&RowMajorBit)
-            dst.copyCoeff(j, i, src);
-          else
-            dst.copyCoeff(i, j, src);
-        }
-    }
+    const int innerSize = dst.innerSize();
+    const int outerSize = dst.outerSize();
+    for(int outer = 0; outer < outerSize; ++outer)
+      for(int inner = 0; inner < innerSize; ++inner)
+        dst.copyCoeffByOuterInner(outer, inner, src);
   }
 };
 
@@ -308,12 +279,10 @@ struct ei_assign_impl<Derived1, Derived2, DefaultTraversal, InnerUnrolling>
 {
   EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src)
   {
-    const bool rowMajor = int(Derived1::Flags)&RowMajorBit;
-    const int innerSize = rowMajor ? Derived1::ColsAtCompileTime : Derived1::RowsAtCompileTime;
     const int outerSize = dst.outerSize();
-    for(int j = 0; j < outerSize; ++j)
-      ei_assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, 0, innerSize>
-        ::run(dst, src, j);
+    for(int outer = 0; outer < outerSize; ++outer)
+      ei_assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, 0, Derived1::InnerSizeAtCompileTime>
+        ::run(dst, src, outer);
   }
 };
 
@@ -354,14 +323,9 @@ struct ei_assign_impl<Derived1, Derived2, InnerVectorizedTraversal, NoUnrolling>
     const int innerSize = dst.innerSize();
     const int outerSize = dst.outerSize();
     const int packetSize = ei_packet_traits<typename Derived1::Scalar>::size;
-    for(int j = 0; j < outerSize; ++j)
-      for(int i = 0; i < innerSize; i+=packetSize)
-      {
-        if(int(Derived1::Flags)&RowMajorBit)
-          dst.template copyPacket<Derived2, Aligned, Aligned>(j, i, src);
-        else
-          dst.template copyPacket<Derived2, Aligned, Aligned>(i, j, src);
-      }
+    for(int outer = 0; outer < outerSize; ++outer)
+      for(int inner = 0; inner < innerSize; inner+=packetSize)
+        dst.template copyPacketByOuterInner<Derived2, Aligned, Aligned>(outer, inner, src);
   }
 };
 
@@ -380,12 +344,10 @@ struct ei_assign_impl<Derived1, Derived2, InnerVectorizedTraversal, InnerUnrolli
 {
   EIGEN_STRONG_INLINE static void run(Derived1 &dst, const Derived2 &src)
   {
-    const bool rowMajor = int(Derived1::Flags)&RowMajorBit;
-    const int innerSize = rowMajor ? Derived1::ColsAtCompileTime : Derived1::RowsAtCompileTime;
     const int outerSize = dst.outerSize();
-    for(int j = 0; j < outerSize; ++j)
-      ei_assign_innervec_InnerUnrolling<Derived1, Derived2, 0, innerSize>
-        ::run(dst, src, j);
+    for(int outer = 0; outer < outerSize; ++outer)
+      ei_assign_innervec_InnerUnrolling<Derived1, Derived2, 0, Derived1::InnerSizeAtCompileTime>
+        ::run(dst, src, outer);
   }
 };
 
@@ -471,36 +433,20 @@ struct ei_assign_impl<Derived1, Derived2, SliceVectorizedTraversal, NoUnrolling>
     int alignedStart = ei_assign_traits<Derived1,Derived2>::DstIsAligned ? 0
                      : ei_first_aligned(&dst.coeffRef(0,0), innerSize);
 
-    for(int i = 0; i < outerSize; ++i)
+    for(int outer = 0; outer < outerSize; ++outer)
     {
       const int alignedEnd = alignedStart + ((innerSize-alignedStart) & ~packetAlignedMask);
-
       // do the non-vectorizable part of the assignment
-      for (int index = 0; index<alignedStart ; ++index)
-      {
-        if(Derived1::Flags&RowMajorBit)
-          dst.copyCoeff(i, index, src);
-        else
-          dst.copyCoeff(index, i, src);
-      }
+      for(int inner = 0; inner<alignedStart ; ++inner)
+        dst.copyCoeffByOuterInner(outer, inner, src);
 
       // do the vectorizable part of the assignment
-      for (int index = alignedStart; index<alignedEnd; index+=packetSize)
-      {
-        if(Derived1::Flags&RowMajorBit)
-          dst.template copyPacket<Derived2, Aligned, Unaligned>(i, index, src);
-        else
-          dst.template copyPacket<Derived2, Aligned, Unaligned>(index, i, src);
-      }
+      for(int inner = alignedStart; inner<alignedEnd; inner+=packetSize)
+        dst.template copyPacketByOuterInner<Derived2, Aligned, Unaligned>(outer, inner, src);
 
       // do the non-vectorizable part of the assignment
-      for (int index = alignedEnd; index<innerSize ; ++index)
-      {
-        if(Derived1::Flags&RowMajorBit)
-          dst.copyCoeff(i, index, src);
-        else
-          dst.copyCoeff(index, i, src);
-      }
+      for(int inner = alignedEnd; inner<innerSize ; ++inner)
+        dst.copyCoeffByOuterInner(outer, inner, src);
 
       alignedStart = std::min<int>((alignedStart+alignedStep)%packetSize, innerSize);
     }
@@ -519,14 +465,6 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>
   EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Derived,OtherDerived)
   EIGEN_STATIC_ASSERT((ei_is_same_type<typename Derived::Scalar, typename OtherDerived::Scalar>::ret),
     YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
-  if(Derived::ColsAtCompileTime == 1)
-  {
-    ei_assert(OtherDerived::RowsAtCompileTime == 1 || other.cols() == 1);
-  }
-  if(Derived::RowsAtCompileTime == 1)
-  {
-    ei_assert(OtherDerived::ColsAtCompileTime == 1 || other.rows() == 1);
-  }
 #ifdef EIGEN_DEBUG_ASSIGN
   ei_assign_traits<Derived, OtherDerived>::debug();
 #endif
diff --git a/Eigen/src/Core/Block.h b/Eigen/src/Core/Block.h
index 8a7aea91f..d3c4dfa99 100644
--- a/Eigen/src/Core/Block.h
+++ b/Eigen/src/Core/Block.h
@@ -80,20 +80,6 @@ struct ei_traits<Block<MatrixType, BlockRows, BlockCols, _DirectAccessStatus> >
   };
 };
 
-template<typename MatrixType, int BlockRows, int BlockCols>
-struct ei_traits<Block<MatrixType, BlockRows, BlockCols, true> > : ei_traits<Block<MatrixType, BlockRows, BlockCols, false> >
-{
-  enum {
-    InnerStrideAtCompileTime =
-        (BlockRows==1 && !(int(MatrixType::Flags)&RowMajorBit))
-        || (BlockCols==1 && (int(MatrixType::Flags)&RowMajorBit))
-        ? MatrixType::OuterStrideAtCompileTime
-        : MatrixType::InnerStrideAtCompileTime,
-    OuterStrideAtCompileTime =
-        (BlockRows==1||BlockCols==1) ? 0 : MatrixType::OuterStrideAtCompileTime
-  };
-};
-
 template<typename MatrixType, int BlockRows, int BlockCols, int _DirectAccessStatus> class Block
   : public MatrixType::template MakeBase< Block<MatrixType, BlockRows, BlockCols, _DirectAccessStatus> >::Type
 {
@@ -114,8 +100,8 @@ template<typename MatrixType, int BlockRows, int BlockCols, int _DirectAccessSta
         // The case a 1x1 matrix seems ambiguous, but the result is the same anyway.
         m_startRow( (BlockRows==1) && (BlockCols==MatrixType::ColsAtCompileTime) ? i : 0),
         m_startCol( (BlockRows==MatrixType::RowsAtCompileTime) && (BlockCols==1) ? i : 0),
-        m_blockRows(matrix.rows()), // if it is a row, then m_blockRows has a fixed-size of 1, so no pb to try to overwrite it
-        m_blockCols(matrix.cols())  // same for m_blockCols
+        m_blockRows(BlockRows==1 ? 1 : matrix.rows()),
+        m_blockCols(BlockCols==1 ? 1 : matrix.cols())
     {
       ei_assert( (i>=0) && (
           ((BlockRows==1) && (BlockCols==MatrixType::ColsAtCompileTime) && i<matrix.rows())
@@ -126,7 +112,7 @@ template<typename MatrixType, int BlockRows, int BlockCols, int _DirectAccessSta
       */
     inline Block(const MatrixType& matrix, int startRow, int startCol)
       : m_matrix(matrix), m_startRow(startRow), m_startCol(startCol),
-        m_blockRows(matrix.rows()), m_blockCols(matrix.cols())
+        m_blockRows(BlockRows), m_blockCols(BlockCols)
     {
       EIGEN_STATIC_ASSERT(RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic,THIS_METHOD_IS_ONLY_FOR_FIXED_SIZE)
       ei_assert(startRow >= 0 && BlockRows >= 1 && startRow + BlockRows <= matrix.rows()
@@ -277,16 +263,15 @@ class Block<MatrixType,BlockRows,BlockCols,HasDirectAccess>
     /** \sa MapBase::innerStride() */
     inline int innerStride() const
     {
-      return (RowsAtCompileTime==1 && !(int(MatrixType::Flags)&RowMajorBit))
-          || (ColsAtCompileTime==1 && (int(MatrixType::Flags)&RowMajorBit))
-          ? m_matrix.outerStride()
-          : m_matrix.innerStride();
+      return RowsAtCompileTime==1 ? m_matrix.colStride()
+           : ColsAtCompileTime==1 ? m_matrix.rowStride()
+           : m_matrix.innerStride();
     }
     
     /** \sa MapBase::outerStride() */
     inline int outerStride() const
     {
-      return IsVectorAtCompileTime ? 0 : m_matrix.outerStride();
+      return IsVectorAtCompileTime ? this->size() : m_matrix.outerStride();
     }
 
   #ifndef __SUNPRO_CC
diff --git a/Eigen/src/Core/Coeffs.h b/Eigen/src/Core/Coeffs.h
index ebfd0c80e..da7b9153f 100644
--- a/Eigen/src/Core/Coeffs.h
+++ b/Eigen/src/Core/Coeffs.h
@@ -25,6 +25,24 @@
 #ifndef EIGEN_COEFFS_H
 #define EIGEN_COEFFS_H
 
+template<typename Derived>
+EIGEN_STRONG_INLINE int DenseBase<Derived>::rowIndexByOuterInner(int outer, int inner)
+{
+  return int(Derived::RowsAtCompileTime) == 1 ? 0
+       : int(Derived::ColsAtCompileTime) == 1 ? inner
+       : int(Derived::Flags)&RowMajorBit ? outer
+       : inner;
+}
+
+template<typename Derived>
+EIGEN_STRONG_INLINE int DenseBase<Derived>::colIndexByOuterInner(int outer, int inner)
+{
+  return int(Derived::ColsAtCompileTime) == 1 ? 0
+       : int(Derived::RowsAtCompileTime) == 1 ? inner
+       : int(Derived::Flags)&RowMajorBit ? inner
+       : outer;
+}
+
 /** Short version: don't use this function, use
   * \link operator()(int,int) const \endlink instead.
   *
@@ -48,6 +66,14 @@ EIGEN_STRONG_INLINE const typename DenseBase<Derived>::CoeffReturnType DenseBase
   return derived().coeff(row, col);
 }
 
+template<typename Derived>
+EIGEN_STRONG_INLINE const typename DenseBase<Derived>::CoeffReturnType DenseBase<Derived>
+  ::coeffByOuterInner(int outer, int inner) const
+{
+  return coeff(rowIndexByOuterInner(outer, inner),
+               colIndexByOuterInner(outer, inner));
+}
+
 /** \returns the coefficient at given the given row and column.
   *
   * \sa operator()(int,int), operator[](int) const
@@ -84,6 +110,14 @@ EIGEN_STRONG_INLINE typename ei_traits<Derived>::Scalar& DenseBase<Derived>
   return derived().coeffRef(row, col);
 }
 
+template<typename Derived>
+EIGEN_STRONG_INLINE typename ei_traits<Derived>::Scalar& DenseBase<Derived>
+  ::coeffRefByOuterInner(int outer, int inner)
+{
+  return coeffRef(rowIndexByOuterInner(outer, inner),
+                  colIndexByOuterInner(outer, inner));
+}
+
 /** \returns a reference to the coefficient at given the given row and column.
   *
   * \sa operator()(int,int) const, operator[](int)
@@ -261,6 +295,15 @@ DenseBase<Derived>::packet(int row, int col) const
   return derived().template packet<LoadMode>(row,col);
 }
 
+template<typename Derived>
+template<int LoadMode>
+EIGEN_STRONG_INLINE typename ei_packet_traits<typename ei_traits<Derived>::Scalar>::type
+DenseBase<Derived>::packetByOuterInner(int outer, int inner) const
+{
+  return packet<LoadMode>(rowIndexByOuterInner(outer, inner),
+                          colIndexByOuterInner(outer, inner));
+}
+
 /** Stores the given packet of coefficients, at the given row and column of this expression. It is your responsibility
   * to ensure that a packet really starts there. This method is only available on expressions having the
   * PacketAccessBit.
@@ -279,6 +322,16 @@ EIGEN_STRONG_INLINE void DenseBase<Derived>::writePacket
   derived().template writePacket<StoreMode>(row,col,x);
 }
 
+template<typename Derived>
+template<int StoreMode>
+EIGEN_STRONG_INLINE void DenseBase<Derived>::writePacketByOuterInner
+(int outer, int inner, const typename ei_packet_traits<typename ei_traits<Derived>::Scalar>::type& x)
+{
+  writePacket<StoreMode>(rowIndexByOuterInner(outer, inner),
+                         colIndexByOuterInner(outer, inner),
+                         x);
+}
+
 /** \returns the packet of coefficients starting at the given index. It is your responsibility
   * to ensure that a packet really starts there. This method is only available on expressions having the
   * PacketAccessBit and the LinearAccessBit.
@@ -346,6 +399,16 @@ EIGEN_STRONG_INLINE void DenseBase<Derived>::copyCoeff(int index, const DenseBas
   derived().coeffRef(index) = other.derived().coeff(index);
 }
 
+template<typename Derived>
+template<typename OtherDerived>
+EIGEN_STRONG_INLINE void DenseBase<Derived>::copyCoeffByOuterInner(int outer, int inner, const DenseBase<OtherDerived>& other)
+{
+  const int row = Derived::rowIndexByOuterInner(outer,inner);
+  const int col = Derived::colIndexByOuterInner(outer,inner);
+  // derived() is important here: copyCoeff() may be reimplemented in Derived!
+  derived().copyCoeff(row, col, other);
+}
+
 /** \internal Copies the packet at position (row,col) of other into *this.
   *
   * This method is overridden in SwapWrapper, allowing swap() assignments to share 99% of their code
@@ -379,6 +442,16 @@ EIGEN_STRONG_INLINE void DenseBase<Derived>::copyPacket(int index, const DenseBa
     other.derived().template packet<LoadMode>(index));
 }
 
+template<typename Derived>
+template<typename OtherDerived, int StoreMode, int LoadMode>
+EIGEN_STRONG_INLINE void DenseBase<Derived>::copyPacketByOuterInner(int outer, int inner, const DenseBase<OtherDerived>& other)
+{
+  const int row = Derived::rowIndexByOuterInner(outer,inner);
+  const int col = Derived::colIndexByOuterInner(outer,inner);
+  // derived() is important here: copyCoeff() may be reimplemented in Derived!
+  derived().copyPacket<OtherDerived, StoreMode, LoadMode>(row, col, other);
+}
+
 template<typename Derived, bool JustReturnZero>
 struct ei_first_aligned_impl
 {
diff --git a/Eigen/src/Core/DenseBase.h b/Eigen/src/Core/DenseBase.h
index d8f789ae0..2078f023b 100644
--- a/Eigen/src/Core/DenseBase.h
+++ b/Eigen/src/Core/DenseBase.h
@@ -124,6 +124,11 @@ template<typename Derived> class DenseBase
           * constructed from this one. See the \ref flags "list of flags".
           */
 
+      IsRowMajor = int(Flags) & RowMajorBit, /**< True if this expression is row major. */
+
+      InnerSizeAtCompileTime = int(IsVectorAtCompileTime) ? SizeAtCompileTime
+                             : int(Flags)&RowMajorBit ? ColsAtCompileTime : RowsAtCompileTime,
+
       CoeffReadCost = ei_traits<Derived>::CoeffReadCost,
         /**< This is a rough measure of how expensive it is to read one coefficient from
           * this expression.
@@ -200,20 +205,64 @@ template<typename Derived> class DenseBase
                 && "DenseBase::resize() does not actually allow to resize.");
     }
 
-    int innerStride() const
+    /** \returns the pointer increment between two consecutive elements.
+      *
+      * \note For vectors, the storage order is ignored. For matrices (non-vectors), we're looking
+      *       at the increment between two consecutive elements within a slice in the inner direction.
+      *
+      * \sa outerStride(), rowStride(), colStride()
+      */
+    inline int innerStride() const
     {
       EIGEN_STATIC_ASSERT(int(Flags)&DirectAccessBit,
                           THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_WITH_DIRECT_MEMORY_ACCESS_SUCH_AS_MAP_OR_PLAIN_MATRICES)
       return derived().innerStride();
     }
 
-    int outerStride() const
+    /** \returns the pointer increment between two consecutive inner slices (for example, between two consecutive columns
+      *          in a column-major matrix).
+      *
+      * \note For vectors, the storage order is ignored, there is only one inner slice, and so this method returns 1.
+      *       For matrices (non-vectors), the notion of inner slice depends on the storage order.
+      *
+      * \sa innerStride(), rowStride(), colStride()
+      */
+    inline int outerStride() const
     {
       EIGEN_STATIC_ASSERT(int(Flags)&DirectAccessBit,
                           THIS_METHOD_IS_ONLY_FOR_EXPRESSIONS_WITH_DIRECT_MEMORY_ACCESS_SUCH_AS_MAP_OR_PLAIN_MATRICES)
       return derived().outerStride();
     }
 
+    inline int stride() const
+    {
+      return IsVectorAtCompileTime ? innerStride() : outerStride();
+    }
+
+    /** \returns the pointer increment between two consecutive rows.
+      *
+      * \sa innerStride(), outerStride(), colStride()
+      */
+    inline int rowStride() const
+    {
+      return ColsAtCompileTime==1 ? innerStride()
+           : RowsAtCompileTime==1 ? outerStride()
+           : IsRowMajor ? outerStride()
+           : innerStride();
+    }
+
+    /** \returns the pointer increment between two consecutive columns.
+      *
+      * \sa innerStride(), outerStride(), rowStride()
+      */
+    inline int colStride() const
+    {
+      return ColsAtCompileTime==1 ? outerStride()
+           : RowsAtCompileTime==1 ? innerStride()
+           : IsRowMajor ? innerStride()
+           : outerStride();
+    }
+
 #ifndef EIGEN_PARSED_BY_DOXYGEN
     /** \internal the return type of coeff()
       */
@@ -269,9 +318,11 @@ template<typename Derived> class DenseBase
     CommaInitializer<Derived> operator<< (const DenseBase<OtherDerived>& other);
 
     const CoeffReturnType coeff(int row, int col) const;
+    const CoeffReturnType coeffByOuterInner(int outer, int inner) const;
     const CoeffReturnType operator()(int row, int col) const;
 
     Scalar& coeffRef(int row, int col);
+    Scalar& coeffRefByOuterInner(int outer, int inner);
     Scalar& operator()(int row, int col);
 
     const CoeffReturnType coeff(int index) const;
@@ -286,17 +337,30 @@ template<typename Derived> class DenseBase
     template<typename OtherDerived>
     void copyCoeff(int row, int col, const DenseBase<OtherDerived>& other);
     template<typename OtherDerived>
+    void copyCoeffByOuterInner(int outer, int inner, const DenseBase<OtherDerived>& other);
+    template<typename OtherDerived>
     void copyCoeff(int index, const DenseBase<OtherDerived>& other);
     template<typename OtherDerived, int StoreMode, int LoadMode>
     void copyPacket(int row, int col, const DenseBase<OtherDerived>& other);
     template<typename OtherDerived, int StoreMode, int LoadMode>
+    void copyPacketByOuterInner(int outer, int inner, const DenseBase<OtherDerived>& other);
+    template<typename OtherDerived, int StoreMode, int LoadMode>
     void copyPacket(int index, const DenseBase<OtherDerived>& other);
+
+  private:
+    static int rowIndexByOuterInner(int outer, int inner);
+    static int colIndexByOuterInner(int outer, int inner);
+  public:
 #endif // not EIGEN_PARSED_BY_DOXYGEN
 
     template<int LoadMode>
     PacketScalar packet(int row, int col) const;
+    template<int LoadMode>
+    PacketScalar packetByOuterInner(int outer, int inner) const;
     template<int StoreMode>
     void writePacket(int row, int col, const PacketScalar& x);
+    template<int StoreMode>
+    void writePacketByOuterInner(int outer, int inner, const PacketScalar& x);
 
     template<int LoadMode>
     PacketScalar packet(int index) const;
diff --git a/Eigen/src/Core/DenseStorageBase.h b/Eigen/src/Core/DenseStorageBase.h
index 04dfb1176..e93e439e6 100644
--- a/Eigen/src/Core/DenseStorageBase.h
+++ b/Eigen/src/Core/DenseStorageBase.h
@@ -75,23 +75,6 @@ class DenseStorageBase : public _Base<Derived>
     EIGEN_STRONG_INLINE int rows() const { return m_storage.rows(); }
     EIGEN_STRONG_INLINE int cols() const { return m_storage.cols(); }
 
-    /** Returns the leading dimension (for matrices) or the increment (for vectors) to be used with data().
-      *
-      * More precisely:
-      *  - for a column major matrix it returns the number of elements between two successive columns
-      *  - for a row major matrix it returns the number of elements between two successive rows
-      *  - for a vector it returns the number of elements between two successive coefficients
-      * This function has to be used together with the MapBase::data() function.
-      *
-      * \sa data() */
-    EIGEN_STRONG_INLINE int stride() const
-    {
-      if(IsVectorAtCompileTime)
-        return 1;
-      else
-        return (Flags & RowMajorBit) ? m_storage.cols() : m_storage.rows();
-    }
-
     EIGEN_STRONG_INLINE const Scalar& coeff(int row, int col) const
     {
       if(Flags & RowMajorBit)
@@ -253,13 +236,13 @@ class DenseStorageBase : public _Base<Derived>
     {
       if(RowsAtCompileTime == 1)
       {
-        ei_assert(other.rows() == 1);
-        resize(1, other.cols());
+        ei_assert(other.rows() == 1 || other.cols() == 1);
+        resize(1, other.size());
       }
       else if(ColsAtCompileTime == 1)
       {
-        ei_assert(other.cols() == 1);
-        resize(other.rows(), 1);
+        ei_assert(other.rows() == 1 || other.cols() == 1);
+        resize(other.size(), 1);
       }
       else resize(other.rows(), other.cols());
     }
diff --git a/Eigen/src/Core/Flagged.h b/Eigen/src/Core/Flagged.h
index 7f42a1e73..9d14aceaa 100644
--- a/Eigen/src/Core/Flagged.h
+++ b/Eigen/src/Core/Flagged.h
@@ -60,7 +60,8 @@ template<typename ExpressionType, unsigned int Added, unsigned int Removed> clas
 
     inline int rows() const { return m_matrix.rows(); }
     inline int cols() const { return m_matrix.cols(); }
-    inline int stride() const { return m_matrix.stride(); }
+    inline int outerStride() const { return m_matrix.outerStride(); }
+    inline int innerStride() const { return m_matrix.innerStride(); }
 
     inline const Scalar coeff(int row, int col) const
     {
diff --git a/Eigen/src/Core/ForceAlignedAccess.h b/Eigen/src/Core/ForceAlignedAccess.h
index 927f43413..300b22329 100644
--- a/Eigen/src/Core/ForceAlignedAccess.h
+++ b/Eigen/src/Core/ForceAlignedAccess.h
@@ -52,7 +52,8 @@ template<typename ExpressionType> class ForceAlignedAccess
 
     inline int rows() const { return m_expression.rows(); }
     inline int cols() const { return m_expression.cols(); }
-    inline int stride() const { return m_expression.stride(); }
+    inline int outerStride() const { return m_expression.outerStride(); }
+    inline int innerStride() const { return m_expression.innerStride(); }
 
     inline const CoeffReturnType coeff(int row, int col) const
     {
diff --git a/Eigen/src/Core/Map.h b/Eigen/src/Core/Map.h
index 432bf1661..d9ccb1b20 100644
--- a/Eigen/src/Core/Map.h
+++ b/Eigen/src/Core/Map.h
@@ -56,14 +56,8 @@ struct ei_traits<Map<MatrixType, Options, StrideType> >
     Flags0 = ei_traits<MatrixType>::Flags,
     Flags1 = ((Options&Aligned)==Aligned ? Flags0 |  AlignedBit
                                          : Flags0 & ~AlignedBit),
-    Flags = int(StrideType::InnerStrideAtCompileTime)==1 ? Flags1 : (Flags1 & ~PacketAccessBit),
-    InnerStrideAtCompileTime = int(StrideType::InnerStrideAtCompileTime) != 0 ? int(StrideType::InnerStrideAtCompileTime) : 1,
-    OuterStrideAtCompileTime =
-        int(StrideType::OuterStrideAtCompileTime != 0) ? int(StrideType::OuterStrideAtCompileTime)
-          : int(MatrixType::IsVectorAtCompileTime) ? int(MatrixType::SizeAtCompileTime)
-          : int(Flags)&RowMajorBit ? int(MatrixType::ColsAtCompileTime)
-          : int(MatrixType::RowsAtCompileTime)
-    };
+    Flags = int(StrideType::InnerStrideAtCompileTime)==1 ? Flags1 : (Flags1 & ~PacketAccessBit)
+  };
 };
 
 template<typename MatrixType, int Options, typename StrideType> class Map
diff --git a/Eigen/src/Core/MapBase.h b/Eigen/src/Core/MapBase.h
index 6bac2ed4c..d735cfc47 100644
--- a/Eigen/src/Core/MapBase.h
+++ b/Eigen/src/Core/MapBase.h
@@ -38,57 +38,22 @@ template<typename Derived, typename Base> class MapBase
   public:
 
     enum {
-      IsRowMajor = (int(ei_traits<Derived>::Flags) & RowMajorBit) ? 1 : 0,
       RowsAtCompileTime = ei_traits<Derived>::RowsAtCompileTime,
       ColsAtCompileTime = ei_traits<Derived>::ColsAtCompileTime,
-      SizeAtCompileTime = Base::SizeAtCompileTime,
-      InnerStrideAtCompileTime = ei_traits<Derived>::InnerStrideAtCompileTime
+      SizeAtCompileTime = Base::SizeAtCompileTime
     };
 
     typedef typename ei_traits<Derived>::Scalar Scalar;
     typedef typename Base::PacketScalar PacketScalar;
     using Base::derived;
+    using Base::innerStride;
+    using Base::outerStride;
+    using Base::rowStride;
+    using Base::colStride;
 
     inline int rows() const { return m_rows.value(); }
     inline int cols() const { return m_cols.value(); }
 
-    /** \returns the pointer increment between two consecutive elements.
-      *
-      * \note For vectors, the storage order is ignored. For matrices (non-vectors), we're looking
-      *       at the increment between two consecutive elements within a slice in the inner direction.
-      *
-      * \sa outerStride(), data(), rowStride(), colStride()
-      */
-    inline int innerStride() const { return derived().innerStride(); }
-
-    /** \returns the pointer increment between two consecutive inner slices (for example, between two consecutive columns
-      *          in a column-major matrix).
-      *
-      * \note For vectors, the storage order is ignored, there is only one inner slice, and so this method returns 1.
-      *       For matrices (non-vectors), the notion of inner slice depends on the storage order.
-      *
-      * \sa innerStride(), data(), rowStride(), colStride()
-      */
-    inline int outerStride() const { return derived().outerStride(); }
-
-    /** \returns the pointer increment between two consecutive rows.
-      *
-      * \sa data(), innerStride(), outerStride(), colStride()
-      */
-    inline int rowStride() const
-    {
-      return (RowsAtCompileTime==1 || IsRowMajor) ? outerStride() : innerStride();
-    }
-
-    /** \returns the pointer increment between two consecutive columns.
-      *
-      * \sa data(), innerStride(), outerStride(), rowStride()
-      */
-    inline int colStride() const
-    {
-      return (RowsAtCompileTime==1 || IsRowMajor) ? innerStride() : outerStride();
-    }
-
     /** Returns a pointer to the first coefficient of the matrix or vector.
       *
       * \note When addressing this data, make sure to honor the strides returned by innerStride() and outerStride().
diff --git a/Eigen/src/Core/Matrix.h b/Eigen/src/Core/Matrix.h
index e011ae8b9..b494b2f00 100644
--- a/Eigen/src/Core/Matrix.h
+++ b/Eigen/src/Core/Matrix.h
@@ -120,10 +120,7 @@ struct ei_traits<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
     MaxRowsAtCompileTime = _MaxRows,
     MaxColsAtCompileTime = _MaxCols,
     Flags = ei_compute_matrix_flags<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::ret,
-    CoeffReadCost = NumTraits<Scalar>::ReadCost,
-    InnerStrideAtCompileTime = 1,
-    OuterStrideAtCompileTime = (RowsAtCompileTime==1||ColsAtCompileTime==1) ? 1
-     : (int(Flags)&RowMajorBit) ? RowsAtCompileTime : ColsAtCompileTime
+    CoeffReadCost = NumTraits<Scalar>::ReadCost
   };
 };
 
diff --git a/Eigen/src/Core/NestByValue.h b/Eigen/src/Core/NestByValue.h
index 9f6d1c0c0..497ce828b 100644
--- a/Eigen/src/Core/NestByValue.h
+++ b/Eigen/src/Core/NestByValue.h
@@ -53,7 +53,8 @@ template<typename ExpressionType> class NestByValue
 
     inline int rows() const { return m_expression.rows(); }
     inline int cols() const { return m_expression.cols(); }
-    inline int stride() const { return m_expression.stride(); }
+    inline int outerStride() const { return m_expression.outerStride(); }
+    inline int innerStride() const { return m_expression.innerStride(); }
 
     inline const CoeffReturnType coeff(int row, int col) const
     {
diff --git a/Eigen/src/Core/Product.h b/Eigen/src/Core/Product.h
index 53277169c..af05773ee 100644
--- a/Eigen/src/Core/Product.h
+++ b/Eigen/src/Core/Product.h
@@ -336,7 +336,7 @@ template<> struct ei_gemv_selector<OnTheRight,ColMajor,true>
     ei_cache_friendly_product_colmajor_times_vector
       <LhsBlasTraits::NeedToConjugate,RhsBlasTraits::NeedToConjugate>(
       dest.size(),
-      &actualLhs.const_cast_derived().coeffRef(0,0), ei_outer_stride_or_outer_size(actualLhs),
+      &actualLhs.const_cast_derived().coeffRef(0,0), actualLhs.stride(),
       actualRhs, actualDest, actualAlpha);
 
     if (!EvalToDest)
@@ -381,7 +381,7 @@ template<> struct ei_gemv_selector<OnTheRight,RowMajor,true>
 
     ei_cache_friendly_product_rowmajor_times_vector
       <LhsBlasTraits::NeedToConjugate,RhsBlasTraits::NeedToConjugate>(
-        &actualLhs.const_cast_derived().coeffRef(0,0), ei_outer_stride_or_outer_size(actualLhs),
+        &actualLhs.const_cast_derived().coeffRef(0,0), actualLhs.stride(),
         rhs_data, prod.rhs().size(), dest, actualAlpha);
 
     if (!DirectlyUseRhs) ei_aligned_stack_delete(Scalar, rhs_data, prod.rhs().size());
diff --git a/Eigen/src/Core/SelfAdjointView.h b/Eigen/src/Core/SelfAdjointView.h
index 6d01ee495..add5a3afb 100644
--- a/Eigen/src/Core/SelfAdjointView.h
+++ b/Eigen/src/Core/SelfAdjointView.h
@@ -75,8 +75,9 @@ template<typename MatrixType, unsigned int UpLo> class SelfAdjointView
 
     inline int rows() const { return m_matrix.rows(); }
     inline int cols() const { return m_matrix.cols(); }
-    inline int stride() const { return m_matrix.stride(); }
-
+    inline int outerStride() const { return m_matrix.outerStride(); }
+    inline int innerStride() const { return m_matrix.innerStride(); }
+    
     /** \sa MatrixBase::coeff()
       * \warning the coordinates must fit into the referenced triangular part
       */
diff --git a/Eigen/src/Core/SelfCwiseBinaryOp.h b/Eigen/src/Core/SelfCwiseBinaryOp.h
index 7ae2e82a4..58aee182d 100644
--- a/Eigen/src/Core/SelfCwiseBinaryOp.h
+++ b/Eigen/src/Core/SelfCwiseBinaryOp.h
@@ -57,7 +57,8 @@ template<typename BinaryOp, typename MatrixType> class SelfCwiseBinaryOp
 
     inline int rows() const { return m_matrix.rows(); }
     inline int cols() const { return m_matrix.cols(); }
-    inline int stride() const { return m_matrix.stride(); }
+    inline int outerStride() const { return m_matrix.outerStride(); }
+    inline int innerStride() const { return m_matrix.innerStride(); }
     inline const Scalar* data() const { return m_matrix.data(); }
 
     // note that this function is needed by assign to correctly align loads/stores
diff --git a/Eigen/src/Core/Stride.h b/Eigen/src/Core/Stride.h
index ba0b19de3..7982035fd 100644
--- a/Eigen/src/Core/Stride.h
+++ b/Eigen/src/Core/Stride.h
@@ -54,20 +54,6 @@ class Stride
     inline int inner() const { return m_inner.value(); }
     inline int outer() const { return m_outer.value(); }
 
-    template<int OtherInnerStrideAtCompileTime, int OtherOuterStrideAtCompileTime>
-    Stride<EIGEN_ENUM_MAX(InnerStrideAtCompileTime, OtherInnerStrideAtCompileTime),
-           EIGEN_ENUM_MAX(OuterStrideAtCompileTime, OtherOuterStrideAtCompileTime)>
-    operator|(const Stride<OtherInnerStrideAtCompileTime, OtherOuterStrideAtCompileTime>& other)
-    {
-      EIGEN_STATIC_ASSERT(!((InnerStrideAtCompileTime && OtherInnerStrideAtCompileTime)
-                         || (OuterStrideAtCompileTime && OtherOuterStrideAtCompileTime)),
-                          YOU_ALREADY_SPECIFIED_THIS_STRIDE)
-      int result_inner = InnerStrideAtCompileTime ? inner() : other.inner();
-      int result_outer = OuterStrideAtCompileTime ? outer() : other.outer();
-      return Stride<EIGEN_ENUM_MAX(InnerStrideAtCompileTime, OtherInnerStrideAtCompileTime),
-                    EIGEN_ENUM_MAX(OuterStrideAtCompileTime, OtherOuterStrideAtCompileTime)>
-                    (result_inner, result_outer);
-    }
   protected:
     ei_int_if_dynamic<InnerStrideAtCompileTime> m_inner;
     ei_int_if_dynamic<OuterStrideAtCompileTime> m_outer;
@@ -91,46 +77,4 @@ class OuterStride : public Stride<0, Value>
     OuterStride(int v) : Base(0,v) {}
 };
 
-template<typename T, bool HasDirectAccess = int(ei_traits<T>::Flags)&DirectAccessBit>
-struct ei_outer_stride_or_outer_size_impl
-{
-  static inline int value(const T& x) { return x.outerStride(); }
-};
-
-template<typename T>
-struct ei_outer_stride_or_outer_size_impl<T, false>
-{
-  static inline int value(const T& x) { return x.outerSize(); }
-};
-
-template<typename T>
-inline int ei_outer_stride_or_outer_size(const T& x)
-{
-  return ei_outer_stride_or_outer_size_impl<T>::value(x);
-}
-
-template<typename T, bool HasDirectAccess = int(ei_traits<typename ei_cleantype<T>::type>::Flags)&DirectAccessBit>
-struct ei_inner_stride_at_compile_time
-{
-  enum { ret = ei_traits<typename ei_cleantype<T>::type>::InnerStrideAtCompileTime };
-};
-
-template<typename T>
-struct ei_inner_stride_at_compile_time<T, false>
-{
-  enum { ret = 1 };
-};
-
-template<typename T, bool HasDirectAccess = int(ei_traits<typename ei_cleantype<T>::type>::Flags)&DirectAccessBit>
-struct ei_outer_stride_at_compile_time
-{
-  enum { ret = ei_traits<typename ei_cleantype<T>::type>::OuterStrideAtCompileTime };
-};
-
-template<typename T>
-struct ei_outer_stride_at_compile_time<T, false>
-{
-  enum { ret = 1 };
-};
-
 #endif // EIGEN_STRIDE_H
diff --git a/Eigen/src/Core/Swap.h b/Eigen/src/Core/Swap.h
index 186268af0..c3c641097 100644
--- a/Eigen/src/Core/Swap.h
+++ b/Eigen/src/Core/Swap.h
@@ -47,7 +47,8 @@ template<typename ExpressionType> class SwapWrapper
 
     inline int rows() const { return m_expression.rows(); }
     inline int cols() const { return m_expression.cols(); }
-    inline int stride() const { return m_expression.stride(); }
+    inline int outerStride() const { return m_expression.outerStride(); }
+    inline int innerStride() const { return m_expression.innerStride(); }
 
     inline Scalar& coeffRef(int row, int col)
     {
@@ -60,7 +61,7 @@ template<typename ExpressionType> class SwapWrapper
     }
 
     template<typename OtherDerived>
-    void copyCoeff(int row, int col, const MatrixBase<OtherDerived>& other)
+    void copyCoeff(int row, int col, const DenseBase<OtherDerived>& other)
     {
       OtherDerived& _other = other.const_cast_derived();
       ei_internal_assert(row >= 0 && row < rows()
@@ -71,7 +72,7 @@ template<typename ExpressionType> class SwapWrapper
     }
 
     template<typename OtherDerived>
-    void copyCoeff(int index, const MatrixBase<OtherDerived>& other)
+    void copyCoeff(int index, const DenseBase<OtherDerived>& other)
     {
       OtherDerived& _other = other.const_cast_derived();
       ei_internal_assert(index >= 0 && index < m_expression.size());
@@ -81,7 +82,7 @@ template<typename ExpressionType> class SwapWrapper
     }
 
     template<typename OtherDerived, int StoreMode, int LoadMode>
-    void copyPacket(int row, int col, const MatrixBase<OtherDerived>& other)
+    void copyPacket(int row, int col, const DenseBase<OtherDerived>& other)
     {
       OtherDerived& _other = other.const_cast_derived();
       ei_internal_assert(row >= 0 && row < rows()
@@ -94,7 +95,7 @@ template<typename ExpressionType> class SwapWrapper
     }
 
     template<typename OtherDerived, int StoreMode, int LoadMode>
-    void copyPacket(int index, const MatrixBase<OtherDerived>& other)
+    void copyPacket(int index, const DenseBase<OtherDerived>& other)
     {
       OtherDerived& _other = other.const_cast_derived();
       ei_internal_assert(index >= 0 && index < m_expression.size());
diff --git a/Eigen/src/Core/Transpose.h b/Eigen/src/Core/Transpose.h
index 753a67ee0..47dae5776 100644
--- a/Eigen/src/Core/Transpose.h
+++ b/Eigen/src/Core/Transpose.h
@@ -80,9 +80,6 @@ template<typename MatrixType> class Transpose
     typename ei_cleantype<typename MatrixType::Nested>::type&
     nestedExpression() { return m_matrix.const_cast_derived(); }
 
-    enum { InnerStrideAtCompileTime = ei_inner_stride_at_compile_time<MatrixType>::ret,
-           OuterStrideAtCompileTime = ei_outer_stride_at_compile_time<MatrixType>::ret };
-
   protected:
     const typename MatrixType::Nested m_matrix;
 };
diff --git a/Eigen/src/Core/TriangularMatrix.h b/Eigen/src/Core/TriangularMatrix.h
index 8bea0aa68..c61a6d7cc 100644
--- a/Eigen/src/Core/TriangularMatrix.h
+++ b/Eigen/src/Core/TriangularMatrix.h
@@ -50,7 +50,8 @@ template<typename Derived> class TriangularBase : public AnyMatrixBase<Derived>
 
     inline int rows() const { return derived().rows(); }
     inline int cols() const { return derived().cols(); }
-    inline int stride() const { return derived().stride(); }
+    inline int outerStride() const { return derived().outerStride(); }
+    inline int innerStride() const { return derived().innerStride(); }
 
     inline Scalar coeff(int row, int col) const  { return derived().coeff(row,col); }
     inline Scalar& coeffRef(int row, int col) { return derived().coeffRef(row,col); }
@@ -165,7 +166,8 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
 
     inline int rows() const { return m_matrix.rows(); }
     inline int cols() const { return m_matrix.cols(); }
-    inline int stride() const { return m_matrix.stride(); }
+    inline int outerStride() const { return m_matrix.outerStride(); }
+    inline int innerStride() const { return m_matrix.innerStride(); }
 
     /** \sa MatrixBase::operator+=() */
     template<typename Other> TriangularView&  operator+=(const Other& other) { return *this = m_matrix + other; }
diff --git a/Eigen/src/Core/VectorBlock.h b/Eigen/src/Core/VectorBlock.h
index cbf97aeb3..5bb7fd35d 100644
--- a/Eigen/src/Core/VectorBlock.h
+++ b/Eigen/src/Core/VectorBlock.h
@@ -86,7 +86,6 @@ template<typename VectorType, int Size> class VectorBlock
              IsColVector ? start : 0, IsColVector ? 0 : start,
              IsColVector ? size  : 1, IsColVector ? 1 : size)
     {
-
       EIGEN_STATIC_ASSERT_VECTOR_ONLY(VectorBlock);
     }
 
diff --git a/Eigen/src/Core/products/GeneralMatrixMatrix.h b/Eigen/src/Core/products/GeneralMatrixMatrix.h
index 03c77cc78..beec17ee4 100644
--- a/Eigen/src/Core/products/GeneralMatrixMatrix.h
+++ b/Eigen/src/Core/products/GeneralMatrixMatrix.h
@@ -147,7 +147,6 @@ class GeneralProduct<Lhs, Rhs, GemmProduct>
 
       const ActualLhsType lhs = LhsBlasTraits::extract(m_lhs);
       const ActualRhsType rhs = RhsBlasTraits::extract(m_rhs);
-      ei_assert(ei_inner_stride_at_compile_time<ActualLhsType>::ret == 1);
 
       Scalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(m_lhs)
                                  * RhsBlasTraits::extractScalarFactor(m_rhs);
@@ -159,9 +158,9 @@ class GeneralProduct<Lhs, Rhs, GemmProduct>
         (Dest::Flags&RowMajorBit) ? RowMajor : ColMajor>
       ::run(
           this->rows(), this->cols(), lhs.cols(),
-          (const Scalar*)&(lhs.const_cast_derived().coeffRef(0,0)), ei_outer_stride_or_outer_size(lhs),
-          (const Scalar*)&(rhs.const_cast_derived().coeffRef(0,0)), ei_outer_stride_or_outer_size(rhs),
-          (Scalar*)&(dst.coeffRef(0,0)), ei_outer_stride_or_outer_size(dst),
+          (const Scalar*)&(lhs.const_cast_derived().coeffRef(0,0)), lhs.stride(),
+          (const Scalar*)&(rhs.const_cast_derived().coeffRef(0,0)), rhs.stride(),
+          (Scalar*)&(dst.coeffRef(0,0)), dst.stride(),
           actualAlpha);
     }
 };
diff --git a/Eigen/src/Core/util/Constants.h b/Eigen/src/Core/util/Constants.h
index 51590b03d..c27c979a6 100644
--- a/Eigen/src/Core/util/Constants.h
+++ b/Eigen/src/Core/util/Constants.h
@@ -140,7 +140,7 @@ const unsigned int LinearAccessBit = 0x10;
   * Means that the underlying array of coefficients can be directly accessed. This means two things.
   * First, references to the coefficients must be available through coeffRef(int, int). This rules out read-only
   * expressions whose coefficients are computed on demand by coeff(int, int). Second, the memory layout of the
-  * array of coefficients must be exactly the natural one suggested by rows(), cols(), stride(), and the RowMajorBit.
+  * array of coefficients must be exactly the natural one suggested by rows(), cols(), outerStride(), innerStride(), and the RowMajorBit.
   * This rules out expressions such as Diagonal, whose coefficients, though referencable, do not have
   * such a regular memory layout.
   */
diff --git a/Eigen/src/Core/util/Memory.h b/Eigen/src/Core/util/Memory.h
index d4920d213..c7b95d334 100644
--- a/Eigen/src/Core/util/Memory.h
+++ b/Eigen/src/Core/util/Memory.h
@@ -232,7 +232,7 @@ inline static Integer ei_first_aligned(const Scalar* array, Integer size)
   enum { PacketSize = ei_packet_traits<Scalar>::size,
          PacketAlignedMask = PacketSize-1
   };
-  
+
   if(PacketSize==1)
   {
     // Either there is no vectorization, or a packet consists of exactly 1 scalar so that all elements